From 7b754a99deb9d2f1b386a7ca074b9b3cf24f100b Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Fri, 19 Apr 2024 14:45:50 +0800
Subject: [PATCH 01/55] added features to download models from the hugging face
 model hub/load local hugging face model and finetune loaded model with
 hugging face dataset

Added features to download models from hugging face model hub/load local hugging face model and finetune loaded model with hugging face dataset. Model loading and fine-tuning can happen both at the initialization stage and after the agent has been initialized (see README in `agentscope/examples/load_finetune_huggingface_model` for details). Major changes to the repo include creating the example script `load_finetune_huggingface_model`, adding a new model wrapper `HuggingFaceWrapper`, and creating a new agent type Finetune_DialogAgent. All changes are done in a new example directory `agentscope/examples/load_finetune_huggingface_model`.
---
 .../load_finetune_huggingface_model/README.md |  52 ++++
 .../huggingface_model.py                      | 236 ++++++++++++++++++
 .../load_finetune_huggingface_model.py        |  43 ++++
 3 files changed, 331 insertions(+)
 create mode 100644 examples/load_finetune_huggingface_model/README.md
 create mode 100644 examples/load_finetune_huggingface_model/huggingface_model.py
 create mode 100644 examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py

diff --git a/examples/load_finetune_huggingface_model/README.md b/examples/load_finetune_huggingface_model/README.md
new file mode 100644
index 000000000..54ae66c94
--- /dev/null
+++ b/examples/load_finetune_huggingface_model/README.md
@@ -0,0 +1,52 @@
+# Multi-Agent Conversation with Custom Model Loading and Fine-Tuning in AgentScope
+
+This example demonstrates how to load and optionally fine-tune a Hugging Face model within a multi-agent conversation setup using AgentScope. The complete code is provided in `load_finetune_huggingface_model.py`.
+
+## Background
+
+In the context of AgentScope, agents are designed to mimic user and assistant roles in a conversation. This setup allows for the integration and testing of different models from the Hugging Face Hub, enhancing their capabilities through fine-tuning with custom datasets.
+
+## Functionality Overview
+
+This example allows you to:
+
+- Set up a user agent and an assistant agent for interactive conversations.
+- Modify the `sys_prompt` to customize the assistant agent's role.
+- Terminate the conversation by entering "exit".
+
+## Advanced Features
+
+Beyond basic conversation setup, the example introduces advanced functionalities:
+
+- Use `dialog_agent.load_model(model_id, local_model_path)` to load a model either from the Hugging Face Model Hub or a local directory.
+- Apply `dialog_agent.fine_tune(data_path)` to fine-tune the model based on your dataset.
+
+## Agent Initialization
+
+When initializing an agent, the following parameters need specification:
+
+- `model_id` (str): Identifier for the model on Hugging Face.
+- `local_model_path` (str): Local path to the model (defaults to loading from Hugging Face if not provided).
+- `data_path` (str): Path to training data (fine-tuning is skipped if not provided).
+- `device` (str): The device (e.g., 'cuda', 'cpu') for model operation, defaulting to 'cuda' if available.
+- `huggingface_token` (from .env file): Token required for models needing authentication from Hugging Face.
+
+## Tested Models
+
+The example is tested using specific Hugging Face models. While it is designed to be flexible, some models may require additional configuration or modification of the provided scripts.
+
+## Prerequisites
+
+Before running this example, ensure you have installed the following packages:
+
+- `transformers`
+- `peft`
+- `python-dotenv`
+- `pytorch`
+- `datasets`
+- `trl`
+
+Additionally, set your Hugging Face token in the `.env` file:
+
+```bash
+python load_finetune_huggingface_model.py
diff --git a/examples/load_finetune_huggingface_model/huggingface_model.py b/examples/load_finetune_huggingface_model/huggingface_model.py
new file mode 100644
index 000000000..00c444638
--- /dev/null
+++ b/examples/load_finetune_huggingface_model/huggingface_model.py
@@ -0,0 +1,236 @@
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from agentscope.agents import DialogAgent
+
+from agentscope.models import ModelWrapperBase, ModelResponse
+from loguru import logger
+
+import torch
+import os
+from dotenv import load_dotenv, find_dotenv
+
+from typing import Optional
+
+class HuggingFaceWrapper(ModelWrapperBase):
+    model_type: str = "huggingface"  # Unique identifier for this model wrapper
+
+    def __init__(self, config_name, model_id, max_length=512, data_path = None, device = None, local_model_path = None, **kwargs):
+        super().__init__(config_name=config_name)
+        self.max_length = max_length  # Set max_length as an attribute
+        self.model_id = model_id
+        relative_path = os.path.join(os.path.dirname(__file__), "../load_finetune_huggingface_model/.env")
+        dotenv_path = os.path.normpath(relative_path)
+        _ = load_dotenv(dotenv_path) # read local .env file
+        huggingface_token  = os.getenv('HUGGINGFACE_TOKEN')
+
+        self.huggingface_token = huggingface_token
+        if device == None:
+            self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+        else:
+            self.device = device
+        try:
+            if local_model_path == None:
+                self.model = AutoModelForCausalLM.from_pretrained(model_id, token = huggingface_token, device_map="auto",)
+                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = huggingface_token)
+                print("load new model")
+            else:
+                self.model = AutoModelForCausalLM.from_pretrained(
+                    local_model_path, local_files_only=True,
+                    device_map="auto"
+                )
+                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = huggingface_token)
+                print("load local model")
+                
+            if data_path != None:
+                self.model = fine_tune(self.model, self.tokenizer, data_path, token = huggingface_token)
+                
+                
+            
+        except Exception as e:
+            logger.error(f"Failed to load model {model_id}: {e}")
+            raise
+
+    def __call__(self, input, **kwargs) -> ModelResponse:
+        try:
+            # Tokenize the input text
+            concatenated_input  = "\n".join([f"{d.get('name', 'System')}: {d['content']}" for d in input])
+            input_ids = self.tokenizer.encode(f"{concatenated_input}\nAssistent: ", return_tensors='pt')
+            # Generate response using the model
+            outputs = self.model.generate(input_ids.to(self.device), max_new_tokens = self.max_length, **kwargs)
+            # Decode the generated tokens to a string
+            generated_text = self.tokenizer.decode(outputs[0][input_ids.shape[1]:], skip_special_tokens=True)
+            
+            return ModelResponse(text=generated_text, raw={'model_id': self.model_id})
+        except Exception as e:
+            logger.error(f"Generation error: {e}")
+            raise
+
+    def load_model(self, model_id, local_model_path = None):
+        """
+        Load a new model for the agent from a local path and update the agent's model.
+        
+        Parameters:
+            local_model_path (str): The file path to the model to be loaded.
+            model_id (str): An identifier for the model on Huggingface.
+        """
+        try:
+            if local_model_path == None:
+                self.model = AutoModelForCausalLM.from_pretrained(model_id, token = self.huggingface_token, device_map="auto",)
+                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
+                print("new model")
+            else:
+                self.model = AutoModelForCausalLM.from_pretrained(
+                    local_model_path, local_files_only=True,
+                    device_map="auto"
+                )
+                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
+                print("local model")
+            
+            
+            # Optionally, log the successful model loading
+            logger.info(f"Successfully loaded new model '{model_id}' from '{local_model_path}'")
+        except Exception as e:
+            # Handle exceptions during model loading, such as file not found or load errors
+            logger.error(f"Failed to load model '{model_id}' from '{local_model_path}': {e}")
+            raise  # Or handle error appropriately
+
+    def fine_tune(self, data_path):
+        """
+        Fine-tune the agent's model using data from the specified path.
+        
+        Parameters:
+            data_path (str): The file path to the training data.
+        """
+        try:
+            self.model = fine_tune(self.model, self.tokenizer, data_path, token = self.huggingface_token)
+            
+            logger.info(f"Successfully fine-tuned model with data from '{data_path}'")
+        except Exception as e:
+            logger.error(f"Failed to fine-tune model with data from '{data_path}': {e}")
+            raise  # Or handle the error appropriately
+
+
+def fine_tune(model, tokenizer, data_path, token):
+    from datasets import load_dataset
+    from datetime import datetime
+    import os
+    import json
+
+    dataset = load_dataset(data_path, token = token)
+
+    from peft import LoraConfig
+
+    lora_config = LoraConfig(
+        r=16,
+        lora_alpha=32,
+        lora_dropout=0.05,
+        bias="none",
+        task_type="CAUSAL_LM"
+    )
+
+    from peft import get_peft_model
+
+    
+
+    model = get_peft_model(model, lora_config)
+
+    from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
+    import transformers
+
+    def formatting_prompts_func(example):
+        output_texts = []
+        for i in range(len(example['conversations'])):
+            text = f"### Question: {example['conversations'][i][0]}\n ### Answer: {example['conversations'][i][1]}"
+            output_texts.append(text)
+        return output_texts
+
+    response_template = " ### Answer:"
+    collator = DataCollatorForCompletionOnlyLM(response_template, tokenizer=tokenizer)
+
+    trainer = SFTTrainer(
+        model,
+        train_dataset=dataset["train"],
+        eval_dataset=dataset["train"],
+        formatting_func=formatting_prompts_func,
+        data_collator=collator,
+        peft_config=lora_config,
+        args=transformers.TrainingArguments(
+            per_device_train_batch_size=1,
+            gradient_accumulation_steps=1,
+            gradient_checkpointing=False,
+            # learning_rate=2e-6,
+            max_steps=10,
+            output_dir="./",
+            optim="paged_adamw_8bit",
+            fp16=True,
+            # num_train_epochs=10.0,
+            logging_steps=1
+        ),
+    )
+
+    print("fine-tuning model")
+
+    trainer.train()
+
+    
+    now = datetime.now()
+    time_string = now.strftime('%Y-%m-%d_%H-%M-%S')
+
+    # Specify the filename
+    log_name = f"{model.config._name_or_path.split('/')[-1]}_{time_string}_log_history.json"
+
+    relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/"+log_name)
+    normalized_path = os.path.normpath(relative_path)
+
+    os.makedirs(os.path.dirname(normalized_path), exist_ok=True)
+
+    # Writing JSON data
+    with open(normalized_path, 'w') as f:
+        json.dump(trainer.state.log_history, f)
+
+    save_name = f"sft_{model.config._name_or_path.split('/')[-1]}_{time_string}"
+    relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/"+save_name)
+    normalized_path = os.path.normpath(relative_path)
+
+    os.makedirs(os.path.dirname(normalized_path), exist_ok=True)
+    # Check if directory exists
+    if not os.path.exists(normalized_path):
+        # If not, create the directory
+        os.makedirs(normalized_path)
+
+    #save model
+    trainer.save_model(normalized_path)
+    # trainer.mode.save_config(save_path)
+
+    
+    return model
+
+
+class Finetune_DialogAgent(DialogAgent):
+    def __init__(self, name: str, sys_prompt: str, model_config_name: str, use_memory: bool = True, memory_config: Optional[dict] = None):
+        super().__init__(name, sys_prompt, model_config_name, use_memory, memory_config)
+
+    def load_model(self, model_id, local_model_path=None):
+        """
+        Load a new model into the agent.
+
+        Parameters:
+            model_id (str): The Hugging Face model ID or a custom identifier.
+            local_model_path (str, optional): Path to a locally saved model.
+        """
+        if hasattr(self.model, "load_model"):
+            self.model.load_model(model_id, local_model_path)
+        else:
+            logger.error("The model wrapper does not support dynamic model loading.")
+
+    def fine_tune(self, data_path):
+        """
+        Fine-tune the agent's underlying model.
+
+        Parameters:
+            data_path (str): The path to the training data.
+        """
+        if hasattr(self.model, "fine_tune"):
+            self.model.fine_tune(data_path)
+            logger.info("Fine-tuning completed successfully.")
+        else:
+            logger.error("The model wrapper does not support fine-tuning.")
diff --git a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
new file mode 100644
index 000000000..b3a6e1b5a
--- /dev/null
+++ b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
@@ -0,0 +1,43 @@
+import agentscope
+from agentscope.agents.user_agent import UserAgent
+from agentscope.pipelines.functional import sequentialpipeline
+from huggingface_model import Finetune_DialogAgent
+
+
+def main() -> None:
+    """A basic conversation demo with a custom model"""
+
+    # Initialize AgentScope with your custom model configuration
+    agentscope.init(
+        model_configs=[
+            {
+                "model_type": "huggingface",
+                "config_name": "my_custom_model",
+                "model_id": "google/gemma-2b-it",  # Or another generative model of your choice
+                # "local_model_path": "/home/zhan1130/agentscope/examples/conversation_basic/sft_gemma-2b-it_2024-04-04_14-22-35",
+                "max_length": 128,
+                "device": "cuda",
+                # "data_path": "GAIR/lima",
+            },
+        ],
+    )
+
+    # Init agents with the custom model
+    dialog_agent = Finetune_DialogAgent(
+        name="Assistant",
+        sys_prompt="You're a helpful assistant.",
+        model_config_name="my_custom_model",  # Use your custom model config name here
+    )
+    
+    dialog_agent.load_model(model_id = "google/gemma-2b-it", local_model_path = None) #load gemma-2b-it from Hugging Face
+    dialog_agent.fine_tune(data_path=  "GAIR/lima") #fine-tune loaded model with lima dataset
+
+    user_agent = UserAgent()
+
+    # Start the conversation between user and assistant
+    x = None
+    while x is None or x.content != "exit":
+        x = sequentialpipeline([dialog_agent, user_agent], x)
+
+if __name__ == "__main__":
+    main()

From ea00db040c01e4639a9c10bf7a3849236c8f8c6d Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Tue, 23 Apr 2024 16:29:16 +0800
Subject: [PATCH 02/55] added customized hyperparameters specification

made customized hyperparameters specification available from `model_configs` for fine-tuning at initialization, or through `fine_tune_config` in `Finetune_DialogAgent`'s `fine_tune` method after initialization
---
 .../load_finetune_huggingface_model/README.md |   6 +-
 .../huggingface_model.py                      | 184 ++++++++++--------
 .../load_finetune_huggingface_model.py        |  25 ++-
 3 files changed, 130 insertions(+), 85 deletions(-)

diff --git a/examples/load_finetune_huggingface_model/README.md b/examples/load_finetune_huggingface_model/README.md
index 54ae66c94..400df461f 100644
--- a/examples/load_finetune_huggingface_model/README.md
+++ b/examples/load_finetune_huggingface_model/README.md
@@ -21,6 +21,8 @@ Beyond basic conversation setup, the example introduces advanced functionalities
 - Use `dialog_agent.load_model(model_id, local_model_path)` to load a model either from the Hugging Face Model Hub or a local directory.
 - Apply `dialog_agent.fine_tune(data_path)` to fine-tune the model based on your dataset.
 
+The default hyperparameters for (SFT) fine-tuning are specified in `agentscope/src/agentscope/models/huggingface_model.py`. For customized hyperparameters, specify them in `model_configs` if the model needs to be fine-tuned at initialization, or specify through `fine_tune_config` in `Finetune_DialogAgent`'s `fine_tune` method after initialization, as shown in the example script `load_finetune_huggingface_model.py`.
+
 ## Agent Initialization
 
 When initializing an agent, the following parameters need specification:
@@ -29,6 +31,7 @@ When initializing an agent, the following parameters need specification:
 - `local_model_path` (str): Local path to the model (defaults to loading from Hugging Face if not provided).
 - `data_path` (str): Path to training data (fine-tuning is skipped if not provided).
 - `device` (str): The device (e.g., 'cuda', 'cpu') for model operation, defaulting to 'cuda' if available.
+- `fine_tune_config` (dict, Optional): A configuration dictionary for fine-tuning the model. It allows specifying hyperparameters and other training options that will be passed to the fine-tuning method. If not provided, default settings will be used. This allows for customization of the fine-tuning process to optimize model performance based on specific requirements.
 - `huggingface_token` (from .env file): Token required for models needing authentication from Hugging Face.
 
 ## Tested Models
@@ -42,11 +45,10 @@ Before running this example, ensure you have installed the following packages:
 - `transformers`
 - `peft`
 - `python-dotenv`
-- `pytorch`
 - `datasets`
 - `trl`
 
 Additionally, set your Hugging Face token in the `.env` file:
 
 ```bash
-python load_finetune_huggingface_model.py
+python load_finetune_huggingface_model.py
\ No newline at end of file
diff --git a/examples/load_finetune_huggingface_model/huggingface_model.py b/examples/load_finetune_huggingface_model/huggingface_model.py
index 00c444638..878bf1f0c 100644
--- a/examples/load_finetune_huggingface_model/huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/huggingface_model.py
@@ -13,10 +13,11 @@
 class HuggingFaceWrapper(ModelWrapperBase):
     model_type: str = "huggingface"  # Unique identifier for this model wrapper
 
-    def __init__(self, config_name, model_id, max_length=512, data_path = None, device = None, local_model_path = None, **kwargs):
+    def __init__(self, config_name, model_id, max_length=512, data_path = None, device = None, local_model_path = None, fine_tune_config=None, **kwargs):
         super().__init__(config_name=config_name)
         self.max_length = max_length  # Set max_length as an attribute
         self.model_id = model_id
+        # relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/.env")
         relative_path = os.path.join(os.path.dirname(__file__), "../load_finetune_huggingface_model/.env")
         dotenv_path = os.path.normpath(relative_path)
         _ = load_dotenv(dotenv_path) # read local .env file
@@ -41,7 +42,7 @@ def __init__(self, config_name, model_id, max_length=512, data_path = None, devi
                 print("load local model")
                 
             if data_path != None:
-                self.model = fine_tune(self.model, self.tokenizer, data_path, token = huggingface_token)
+                self.model = self.fine_tune_training(self.model, self.tokenizer, data_path,  token = self.huggingface_token, fine_tune_config = fine_tune_config)
                 
                 
             
@@ -93,7 +94,7 @@ def load_model(self, model_id, local_model_path = None):
             logger.error(f"Failed to load model '{model_id}' from '{local_model_path}': {e}")
             raise  # Or handle error appropriately
 
-    def fine_tune(self, data_path):
+    def fine_tune(self, data_path, fine_tune_config=None):
         """
         Fine-tune the agent's model using data from the specified path.
         
@@ -101,7 +102,7 @@ def fine_tune(self, data_path):
             data_path (str): The file path to the training data.
         """
         try:
-            self.model = fine_tune(self.model, self.tokenizer, data_path, token = self.huggingface_token)
+            self.model = self.fine_tune_training(self.model, self.tokenizer, data_path, token = self.huggingface_token, fine_tune_config = fine_tune_config)
             
             logger.info(f"Successfully fine-tuned model with data from '{data_path}'")
         except Exception as e:
@@ -109,103 +110,128 @@ def fine_tune(self, data_path):
             raise  # Or handle the error appropriately
 
 
-def fine_tune(model, tokenizer, data_path, token):
-    from datasets import load_dataset
-    from datetime import datetime
-    import os
-    import json
+    def fine_tune_training(self, model, tokenizer, data_path, token, fine_tune_config=None):
+        from datasets import load_dataset
+        from datetime import datetime
+        import os
+        import json
 
-    dataset = load_dataset(data_path, token = token)
+        dataset = load_dataset(data_path, token = token)
 
-    from peft import LoraConfig
+        from peft import LoraConfig
 
-    lora_config = LoraConfig(
-        r=16,
-        lora_alpha=32,
-        lora_dropout=0.05,
-        bias="none",
-        task_type="CAUSAL_LM"
-    )
+        lora_config_default = {
+            "r": 16,
+            "lora_alpha": 32,
+            "lora_dropout": 0.05,
+            "bias": "none",
+            "task_type": "CAUSAL_LM"
+        }
 
-    from peft import get_peft_model
+        if fine_tune_config is not None:
+            if fine_tune_config['lora_config'] is not None:
+                lora_config_default.update(fine_tune_config['lora_config'])
 
-    
 
-    model = get_peft_model(model, lora_config)
+        training_defaults = {
+            "per_device_train_batch_size": 1,
+            "gradient_accumulation_steps": 1,
+            "gradient_checkpointing": False,
+            "max_steps": 10,
+            "output_dir": "./",
+            "optim": "paged_adamw_8bit",
+            "fp16": True,
+            "logging_steps": 1,
+            # "learning_rate": 2e-6,
+            # "num_train_epochs": 10.0,
+        }
 
-    from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
-    import transformers
+        if fine_tune_config is not None:
+            if fine_tune_config['training_args'] is not None:
+                training_defaults.update(fine_tune_config['training_args'])
 
-    def formatting_prompts_func(example):
-        output_texts = []
-        for i in range(len(example['conversations'])):
-            text = f"### Question: {example['conversations'][i][0]}\n ### Answer: {example['conversations'][i][1]}"
-            output_texts.append(text)
-        return output_texts
+        from peft import get_peft_model
 
-    response_template = " ### Answer:"
-    collator = DataCollatorForCompletionOnlyLM(response_template, tokenizer=tokenizer)
+        
+        lora_config = LoraConfig(**lora_config_default)
+        model = get_peft_model(model, lora_config)
+
+        from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
+        import transformers
 
-    trainer = SFTTrainer(
-        model,
-        train_dataset=dataset["train"],
-        eval_dataset=dataset["train"],
-        formatting_func=formatting_prompts_func,
-        data_collator=collator,
-        peft_config=lora_config,
-        args=transformers.TrainingArguments(
-            per_device_train_batch_size=1,
-            gradient_accumulation_steps=1,
-            gradient_checkpointing=False,
-            # learning_rate=2e-6,
-            max_steps=10,
-            output_dir="./",
-            optim="paged_adamw_8bit",
-            fp16=True,
-            # num_train_epochs=10.0,
-            logging_steps=1
-        ),
-    )
+        def formatting_prompts_func(example):
+            output_texts = []
+            for i in range(len(example['conversations'])):
+                text = f"### Question: {example['conversations'][i][0]}\n ### Answer: {example['conversations'][i][1]}"
+                output_texts.append(text)
+            return output_texts
 
-    print("fine-tuning model")
+        response_template = " ### Answer:"
+        collator = DataCollatorForCompletionOnlyLM(response_template, tokenizer=tokenizer)
 
-    trainer.train()
+        trainer_args = transformers.TrainingArguments(**training_defaults)
 
-    
-    now = datetime.now()
-    time_string = now.strftime('%Y-%m-%d_%H-%M-%S')
+        trainer = SFTTrainer(
+            model,
+            train_dataset=dataset["train"],
+            eval_dataset=dataset["train"],
+            formatting_func=formatting_prompts_func,
+            data_collator=collator,
+            peft_config=lora_config,
+            args=trainer_args,
+        )
 
-    # Specify the filename
-    log_name = f"{model.config._name_or_path.split('/')[-1]}_{time_string}_log_history.json"
+        print("fine-tuning model")
 
-    relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/"+log_name)
-    normalized_path = os.path.normpath(relative_path)
+        trainer.train()
+
+        
+        now = datetime.now()
+        time_string = now.strftime('%Y-%m-%d_%H-%M-%S')
 
-    os.makedirs(os.path.dirname(normalized_path), exist_ok=True)
+        # Specify the filename
+        log_name = f"{model.config._name_or_path.split('/')[-1]}_{time_string}_log_history.json"
 
-    # Writing JSON data
-    with open(normalized_path, 'w') as f:
-        json.dump(trainer.state.log_history, f)
+        relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/"+log_name)
+        normalized_path = os.path.normpath(relative_path)
 
-    save_name = f"sft_{model.config._name_or_path.split('/')[-1]}_{time_string}"
-    relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/"+save_name)
-    normalized_path = os.path.normpath(relative_path)
+        os.makedirs(os.path.dirname(normalized_path), exist_ok=True)
 
-    os.makedirs(os.path.dirname(normalized_path), exist_ok=True)
-    # Check if directory exists
-    if not os.path.exists(normalized_path):
-        # If not, create the directory
-        os.makedirs(normalized_path)
+        # Writing JSON data
+        with open(normalized_path, 'w') as f:
+            json.dump(trainer.state.log_history, f)
 
-    #save model
-    trainer.save_model(normalized_path)
-    # trainer.mode.save_config(save_path)
+        save_name = f"sft_{model.config._name_or_path.split('/')[-1]}_{time_string}"
+        relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/"+save_name)
+        normalized_path = os.path.normpath(relative_path)
 
-    
-    return model
+        os.makedirs(os.path.dirname(normalized_path), exist_ok=True)
+        # Check if directory exists
+        if not os.path.exists(normalized_path):
+            # If not, create the directory
+            os.makedirs(normalized_path)
+
+        #save model
+        trainer.save_model(normalized_path)
+        # trainer.mode.save_config(save_path)
+
+        
+        return model
 
 
 class Finetune_DialogAgent(DialogAgent):
+    """
+    A dialog agent capable of fine-tuning its underlying model based on provided data.
+
+    Inherits from DialogAgent and adds functionality for fine-tuning with custom hyperparameters.
+
+    Parameters:
+        name (str): Name of the agent.
+        sys_prompt (str): System prompt or description of the agent's role.
+        model_config_name (str): The configuration name for the underlying model.
+        use_memory (bool, optional): Whether to use memory for the agent. Defaults to True.
+        memory_config (dict, Optional): Configuration for the agent's memory. Defaults to None.
+    """
     def __init__(self, name: str, sys_prompt: str, model_config_name: str, use_memory: bool = True, memory_config: Optional[dict] = None):
         super().__init__(name, sys_prompt, model_config_name, use_memory, memory_config)
 
@@ -222,7 +248,7 @@ def load_model(self, model_id, local_model_path=None):
         else:
             logger.error("The model wrapper does not support dynamic model loading.")
 
-    def fine_tune(self, data_path):
+    def fine_tune(self, data_path, fine_tune_config=None):
         """
         Fine-tune the agent's underlying model.
 
@@ -230,7 +256,7 @@ def fine_tune(self, data_path):
             data_path (str): The path to the training data.
         """
         if hasattr(self.model, "fine_tune"):
-            self.model.fine_tune(data_path)
+            self.model.fine_tune(data_path, fine_tune_config)
             logger.info("Fine-tuning completed successfully.")
         else:
             logger.error("The model wrapper does not support fine-tuning.")
diff --git a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
index b3a6e1b5a..69965a938 100644
--- a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
@@ -8,16 +8,27 @@ def main() -> None:
     """A basic conversation demo with a custom model"""
 
     # Initialize AgentScope with your custom model configuration
+    
+
     agentscope.init(
         model_configs=[
             {
                 "model_type": "huggingface",
                 "config_name": "my_custom_model",
                 "model_id": "google/gemma-2b-it",  # Or another generative model of your choice
-                # "local_model_path": "/home/zhan1130/agentscope/examples/conversation_basic/sft_gemma-2b-it_2024-04-04_14-22-35",
+                # "local_model_path": # Specify your local model path
                 "max_length": 128,
                 "device": "cuda",
-                # "data_path": "GAIR/lima",
+
+                # "data_path": "GAIR/lima", # Specify a Hugging Face data path if you wish to finetune the model from the start
+
+                # fine_tune_config (Optional): Configuration for fine-tuning the model. This dictionary
+                # can include hyperparameters and other training options that
+                # will be passed to the fine-tuning method. Defaults to None.
+                # "fine_tune_config":{
+                # "lora_config": {"r": 20, "lora_alpha": 40},
+                # "training_args": {"max_steps": 20, "logging_steps": 2}
+                # }
             },
         ],
     )
@@ -29,8 +40,14 @@ def main() -> None:
         model_config_name="my_custom_model",  # Use your custom model config name here
     )
     
-    dialog_agent.load_model(model_id = "google/gemma-2b-it", local_model_path = None) #load gemma-2b-it from Hugging Face
-    dialog_agent.fine_tune(data_path=  "GAIR/lima") #fine-tune loaded model with lima dataset
+    dialog_agent.load_model(model_id = "google/gemma-2b", local_model_path = None) #load gemma-2b-it from Hugging Face
+    # dialog_agent.fine_tune(data_path=  "GAIR/lima") #fine-tune loaded model with lima dataset with default hyperparameters
+    
+    #fine-tune loaded model with lima dataset with customized hyperparameters (`fine_tune_config` argument is optional. Defaults to None.)
+    dialog_agent.fine_tune("GAIR/lima", fine_tune_config ={
+    "lora_config": {"r": 24, "lora_alpha": 48},
+    "training_args": {"max_steps": 30, "logging_steps": 3}
+    })
 
     user_agent = UserAgent()
 

From 3e8c46839233d7f061a8f036a1c6b90484c7b034 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 25 Apr 2024 13:15:23 +0800
Subject: [PATCH 03/55] added docstring and made changes in accordance with the
 comments

---
 .../huggingface_model.py                      | 115 +++++++++++++++---
 .../load_finetune_huggingface_model.py        |  13 +-
 2 files changed, 102 insertions(+), 26 deletions(-)

diff --git a/examples/load_finetune_huggingface_model/huggingface_model.py b/examples/load_finetune_huggingface_model/huggingface_model.py
index 878bf1f0c..fa1b178c4 100644
--- a/examples/load_finetune_huggingface_model/huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/huggingface_model.py
@@ -11,34 +11,48 @@
 from typing import Optional
 
 class HuggingFaceWrapper(ModelWrapperBase):
+    """Wrapper for a Hugging Face transformer model.
+
+    This class is responsible for loading and fine-tuning pre-trained models from the Hugging Face library.
+    """
     model_type: str = "huggingface"  # Unique identifier for this model wrapper
 
     def __init__(self, config_name, model_id, max_length=512, data_path = None, device = None, local_model_path = None, fine_tune_config=None, **kwargs):
+        """Initializes the HuggingFaceWrapper with the given configuration.
+
+        Arguments:
+            config_name (str): Configuration name for model setup.
+            model_id (str): Identifier for the pre-trained model on Hugging Face.
+            max_length (int): Maximum sequence length for the model output per reply. Defaults to 512.
+            data_path (str, optional): Path to the dataset for fine-tuning the model.
+            device (torch.device, optional): Device to run the model on. Will default to GPU if available.
+            local_model_path (str, optional): Local file path to a pre-trained model.
+            fine_tune_config (dict, optional): Configuration for fine-tuning the model.
+            **kwargs: Additional keyword arguments.
+        """
         super().__init__(config_name=config_name)
         self.max_length = max_length  # Set max_length as an attribute
         self.model_id = model_id
-        # relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/.env")
         relative_path = os.path.join(os.path.dirname(__file__), "../load_finetune_huggingface_model/.env")
         dotenv_path = os.path.normpath(relative_path)
         _ = load_dotenv(dotenv_path) # read local .env file
-        huggingface_token  = os.getenv('HUGGINGFACE_TOKEN')
+        self.huggingface_token  = os.getenv('HUGGINGFACE_TOKEN')
 
-        self.huggingface_token = huggingface_token
         if device == None:
             self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
         else:
             self.device = device
         try:
             if local_model_path == None:
-                self.model = AutoModelForCausalLM.from_pretrained(model_id, token = huggingface_token, device_map="auto",)
-                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = huggingface_token)
+                self.model = AutoModelForCausalLM.from_pretrained(model_id, token = self.huggingface_token, device_map="auto",)
+                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
                 print("load new model")
             else:
                 self.model = AutoModelForCausalLM.from_pretrained(
                     local_model_path, local_files_only=True,
                     device_map="auto"
                 )
-                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = huggingface_token)
+                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
                 print("load local model")
                 
             if data_path != None:
@@ -51,6 +65,22 @@ def __init__(self, config_name, model_id, max_length=512, data_path = None, devi
             raise
 
     def __call__(self, input, **kwargs) -> ModelResponse:
+        """Process the input data to generate a response from the model.
+
+        This method tokenizes the input text, generates a response using the model,
+        and then decodes the generated tokens into a string.
+
+        Arguments:
+            input (list): A list of dictionaries where each dictionary contains 'name', 'role' and 'content' keys and their respective values.
+            **kwargs: Additional keyword arguments for the model's generate function.
+
+        Returns:
+            ModelResponse: An object containing the generated text and raw model output.
+
+        Raises:
+            Exception: If an error occurs during text generation.
+        """
+
         try:
             # Tokenize the input text
             concatenated_input  = "\n".join([f"{d.get('name', 'System')}: {d['content']}" for d in input])
@@ -60,7 +90,7 @@ def __call__(self, input, **kwargs) -> ModelResponse:
             # Decode the generated tokens to a string
             generated_text = self.tokenizer.decode(outputs[0][input_ids.shape[1]:], skip_special_tokens=True)
             
-            return ModelResponse(text=generated_text, raw={'model_id': self.model_id})
+            return ModelResponse(text=generated_text, raw = outputs)
         except Exception as e:
             logger.error(f"Generation error: {e}")
             raise
@@ -69,10 +99,14 @@ def load_model(self, model_id, local_model_path = None):
         """
         Load a new model for the agent from a local path and update the agent's model.
         
-        Parameters:
+        Arguments:
             local_model_path (str): The file path to the model to be loaded.
             model_id (str): An identifier for the model on Huggingface.
+        
+        Raises:
+            Exception: If the model cannot be loaded from the given path or identifier. Possible reasons include file not found, incorrect model ID, or network issues while fetching the model.
         """
+
         try:
             if local_model_path == None:
                 self.model = AutoModelForCausalLM.from_pretrained(model_id, token = self.huggingface_token, device_map="auto",)
@@ -98,8 +132,11 @@ def fine_tune(self, data_path, fine_tune_config=None):
         """
         Fine-tune the agent's model using data from the specified path.
         
-        Parameters:
-            data_path (str): The file path to the training data.
+        Arguments:
+            data_path (str): The file path to the training data from Hugging Face.
+            
+        Raises:
+            Exception: If the fine-tuning process fails. This could be due to issues with the data path, configuration parameters, or internal errors during the training process.
         """
         try:
             self.model = self.fine_tune_training(self.model, self.tokenizer, data_path, token = self.huggingface_token, fine_tune_config = fine_tune_config)
@@ -111,6 +148,28 @@ def fine_tune(self, data_path, fine_tune_config=None):
 
 
     def fine_tune_training(self, model, tokenizer, data_path, token, fine_tune_config=None):
+        
+        """
+        The actual method that handles training and fine-tuning the model on the dataset specified by the data_path using a given tokenizer.
+
+        Arguments:
+            model (AutoModelForCausalLM): The pre-trained causal language model from Hugging Face's transformers.
+            tokenizer (AutoTokenizer): The tokenizer corresponding to the pre-trained model.
+            data_path (str): The file path or dataset identifier to load the dataset from Hugging Face.
+            token (str): The authentication token for Hugging Face.
+            fine_tune_config (dict, optional): Configuration options for fine-tuning the model, including LoRA and training arguments.
+
+        Returns:
+            AutoModelForCausalLM: The fine-tuned language model.
+
+        Raises:
+            Exception: Raises an exception if the dataset loading or fine-tuning process fails.
+
+        Note:
+            This method updates the model in place and also logs the fine-tuning process.
+            It utilizes the LoRA configuration and custom training arguments to adapt the pre-trained model to the specific dataset.
+            The training log and trained model are saved in the same directory with the specific timestamp at saving time as part of the log/model fodler name. 
+        """
         from datasets import load_dataset
         from datetime import datetime
         import os
@@ -224,25 +283,39 @@ class Finetune_DialogAgent(DialogAgent):
     A dialog agent capable of fine-tuning its underlying model based on provided data.
 
     Inherits from DialogAgent and adds functionality for fine-tuning with custom hyperparameters.
-
-    Parameters:
-        name (str): Name of the agent.
-        sys_prompt (str): System prompt or description of the agent's role.
-        model_config_name (str): The configuration name for the underlying model.
-        use_memory (bool, optional): Whether to use memory for the agent. Defaults to True.
-        memory_config (dict, Optional): Configuration for the agent's memory. Defaults to None.
     """
+
     def __init__(self, name: str, sys_prompt: str, model_config_name: str, use_memory: bool = True, memory_config: Optional[dict] = None):
+        """
+        Initializes a new Finetune_DialogAgent with specified configuration.
+
+        Arguments:
+            name (str): Name of the agent.
+            sys_prompt (str): System prompt or description of the agent's role.
+            model_config_name (str): The configuration name for the underlying model.
+            use_memory (bool, optional): Indicates whether to utilize memory features. Defaults to True.
+            memory_config (dict, optional): Configuration for memory functionalities if `use_memory` is True.
+        
+        Note:
+            Refer to `class DialogAgent(AgentBase)` for more information.
+        """
+        
         super().__init__(name, sys_prompt, model_config_name, use_memory, memory_config)
+        
+        
 
     def load_model(self, model_id, local_model_path=None):
         """
         Load a new model into the agent.
 
-        Parameters:
+        Arguments:
             model_id (str): The Hugging Face model ID or a custom identifier.
             local_model_path (str, optional): Path to a locally saved model.
+        
+        Raises:
+            Exception: If the model loading process fails or if the model wrapper does not support dynamic loading.
         """
+
         if hasattr(self.model, "load_model"):
             self.model.load_model(model_id, local_model_path)
         else:
@@ -252,9 +325,13 @@ def fine_tune(self, data_path, fine_tune_config=None):
         """
         Fine-tune the agent's underlying model.
 
-        Parameters:
+        Arguments:
             data_path (str): The path to the training data.
+        
+        Raises:
+            Exception: If fine-tuning fails or if the model wrapper does not support fine-tuning.
         """
+
         if hasattr(self.model, "fine_tune"):
             self.model.fine_tune(data_path, fine_tune_config)
             logger.info("Fine-tuning completed successfully.")
diff --git a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
index 69965a938..7a38654d4 100644
--- a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
@@ -19,16 +19,15 @@ def main() -> None:
                 # "local_model_path": # Specify your local model path
                 "max_length": 128,
                 "device": "cuda",
-
-                # "data_path": "GAIR/lima", # Specify a Hugging Face data path if you wish to finetune the model from the start
+                "data_path": "GAIR/lima", # Specify a Hugging Face data path if you wish to finetune the model from the start
 
                 # fine_tune_config (Optional): Configuration for fine-tuning the model. This dictionary
                 # can include hyperparameters and other training options that
                 # will be passed to the fine-tuning method. Defaults to None.
-                # "fine_tune_config":{
-                # "lora_config": {"r": 20, "lora_alpha": 40},
-                # "training_args": {"max_steps": 20, "logging_steps": 2}
-                # }
+                "fine_tune_config":{
+                "lora_config": {"r": 20, "lora_alpha": 40},
+                "training_args": {"max_steps": 20, "logging_steps": 2}
+                }
             },
         ],
     )
@@ -40,7 +39,7 @@ def main() -> None:
         model_config_name="my_custom_model",  # Use your custom model config name here
     )
     
-    dialog_agent.load_model(model_id = "google/gemma-2b", local_model_path = None) #load gemma-2b-it from Hugging Face
+    dialog_agent.load_model(model_id = "google/gemma-2b-it", local_model_path = None) #load gemma-2b-it from Hugging Face
     # dialog_agent.fine_tune(data_path=  "GAIR/lima") #fine-tune loaded model with lima dataset with default hyperparameters
     
     #fine-tune loaded model with lima dataset with customized hyperparameters (`fine_tune_config` argument is optional. Defaults to None.)

From 10a9870b0dd70cda86ab6f2d41f79e49c68d4a04 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 25 Apr 2024 17:15:08 +0800
Subject: [PATCH 04/55] decoupled model loading and tokenizer loading. Now can
 load tokenizer from local.

---
 .../huggingface_model.py                      | 112 +++++++++++-------
 .../load_finetune_huggingface_model.py        |   9 +-
 2 files changed, 75 insertions(+), 46 deletions(-)

diff --git a/examples/load_finetune_huggingface_model/huggingface_model.py b/examples/load_finetune_huggingface_model/huggingface_model.py
index fa1b178c4..8c48f6ec0 100644
--- a/examples/load_finetune_huggingface_model/huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/huggingface_model.py
@@ -17,7 +17,7 @@ class HuggingFaceWrapper(ModelWrapperBase):
     """
     model_type: str = "huggingface"  # Unique identifier for this model wrapper
 
-    def __init__(self, config_name, model_id, max_length=512, data_path = None, device = None, local_model_path = None, fine_tune_config=None, **kwargs):
+    def __init__(self, config_name, model_id = None, max_length=512, data_path = None, device = None, local_model_path = None, local_tokenizer_path = None, fine_tune_config=None, **kwargs):
         """Initializes the HuggingFaceWrapper with the given configuration.
 
         Arguments:
@@ -42,28 +42,13 @@ def __init__(self, config_name, model_id, max_length=512, data_path = None, devi
             self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
         else:
             self.device = device
-        try:
-            if local_model_path == None:
-                self.model = AutoModelForCausalLM.from_pretrained(model_id, token = self.huggingface_token, device_map="auto",)
-                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
-                print("load new model")
-            else:
-                self.model = AutoModelForCausalLM.from_pretrained(
-                    local_model_path, local_files_only=True,
-                    device_map="auto"
-                )
-                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
-                print("load local model")
-                
-            if data_path != None:
-                self.model = self.fine_tune_training(self.model, self.tokenizer, data_path,  token = self.huggingface_token, fine_tune_config = fine_tune_config)
-                
+        
+        self.load_model(model_id, local_model_path = local_model_path)
+        self.load_tokenizer(model_id, local_tokenizer_path = local_tokenizer_path)
+        
+        if data_path != None:
+            self.model = self.fine_tune_training(self.model, self.tokenizer, data_path,  token = self.huggingface_token, fine_tune_config = fine_tune_config)
                 
-            
-        except Exception as e:
-            logger.error(f"Failed to load model {model_id}: {e}")
-            raise
-
     def __call__(self, input, **kwargs) -> ModelResponse:
         """Process the input data to generate a response from the model.
 
@@ -110,23 +95,47 @@ def load_model(self, model_id, local_model_path = None):
         try:
             if local_model_path == None:
                 self.model = AutoModelForCausalLM.from_pretrained(model_id, token = self.huggingface_token, device_map="auto",)
-                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
                 print("new model")
             else:
                 self.model = AutoModelForCausalLM.from_pretrained(
                     local_model_path, local_files_only=True,
                     device_map="auto"
                 )
-                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
                 print("local model")
             
-            
-            # Optionally, log the successful model loading
+            #log the successful model loading
             logger.info(f"Successfully loaded new model '{model_id}' from '{local_model_path}'")
         except Exception as e:
             # Handle exceptions during model loading, such as file not found or load errors
             logger.error(f"Failed to load model '{model_id}' from '{local_model_path}': {e}")
             raise  # Or handle error appropriately
+    
+    def load_tokenizer(self, model_id, local_tokenizer_path = None):
+        """
+        Load the tokenizer from a local path.
+        
+        Arguments:
+            local_tokenizer_path (str): The file path to the tokenizer to be loaded.
+            model_id (str): An identifier for the model on Huggingface.
+        
+        Raises:
+            Exception: If the tokenizer cannot be loaded from the given path or identifier. Possible reasons include file not found, incorrect model ID, or network issues while fetching the tokenizer.
+        """
+
+        try:
+            if local_tokenizer_path == None:
+                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
+                print("new tokenizer")
+            else:
+                self.tokenizer = AutoTokenizer.from_pretrained(local_tokenizer_path)
+                print("local tokenizer")
+            
+            #log the successful tokenizer loading
+            logger.info(f"Successfully loaded new tokenizer for model '{model_id}' from '{local_tokenizer_path}'")
+        except Exception as e:
+            # Handle exceptions during model loading, such as file not found or load errors
+            logger.error(f"Failed to load tokenizer for model '{model_id}' from '{local_tokenizer_path}': {e}")
+            raise  # Or handle error appropriately
 
     def fine_tune(self, data_path, fine_tune_config=None):
         """
@@ -250,29 +259,29 @@ def formatting_prompts_func(example):
 
         # Specify the filename
         log_name = f"{model.config._name_or_path.split('/')[-1]}_{time_string}_log_history.json"
-
-        relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/"+log_name)
-        normalized_path = os.path.normpath(relative_path)
-
-        os.makedirs(os.path.dirname(normalized_path), exist_ok=True)
+        log_path = os.path.join(os.path.dirname(__file__), log_name)
 
         # Writing JSON data
-        with open(normalized_path, 'w') as f:
+        with open(log_path, 'w') as f:
             json.dump(trainer.state.log_history, f)
 
-        save_name = f"sft_{model.config._name_or_path.split('/')[-1]}_{time_string}"
-        relative_path = os.path.join(os.path.dirname(__file__), "../../../examples/load_finetune_huggingface_model/"+save_name)
-        normalized_path = os.path.normpath(relative_path)
+        
 
-        os.makedirs(os.path.dirname(normalized_path), exist_ok=True)
+        # os.makedirs(os.path.dirname(model_path), exist_ok=True)
         # Check if directory exists
-        if not os.path.exists(normalized_path):
-            # If not, create the directory
-            os.makedirs(normalized_path)
+        # if not os.path.exists(model_path):
+        #     # If not, create the directory
+        #     os.makedirs(model_path)
 
         #save model
-        trainer.save_model(normalized_path)
-        # trainer.mode.save_config(save_path)
+        model_name = f"sft_{model.config._name_or_path.split('/')[-1]}_{time_string}"
+        model_path = os.path.join(os.path.dirname(__file__), model_name)
+        trainer.save_model(model_path)
+
+        #save tokenizer
+        tokenizer_name = f"sft_{model.config._name_or_path.split('/')[-1]}_tokenizer_{time_string}"
+        tokenizer_path = os.path.join(os.path.dirname(__file__), tokenizer_name)
+        tokenizer.save_pretrained(tokenizer_path)
 
         
         return model
@@ -304,12 +313,12 @@ def __init__(self, name: str, sys_prompt: str, model_config_name: str, use_memor
         
         
 
-    def load_model(self, model_id, local_model_path=None):
+    def load_model(self, model_id = None, local_model_path=None):
         """
         Load a new model into the agent.
 
         Arguments:
-            model_id (str): The Hugging Face model ID or a custom identifier.
+            model_id (str): The Hugging Face model ID or a custom identifier. Needed if loading model from Hugging Face.
             local_model_path (str, optional): Path to a locally saved model.
         
         Raises:
@@ -321,6 +330,23 @@ def load_model(self, model_id, local_model_path=None):
         else:
             logger.error("The model wrapper does not support dynamic model loading.")
 
+    def load_tokenizer(self, model_id = None, local_tokenizer_path=None):
+        """
+        Load a new tokenizer for the agent.
+
+        Arguments:
+            model_id (str): The Hugging Face model ID or a custom identifier. Needed if loading tokenizer from Hugging Face.
+            local_tokenizer_path (str, optional): Path to a locally saved tokenizer.
+        
+        Raises:
+            Exception: If the model tokenizer process fails or if the model wrapper does not support dynamic loading.
+        """
+
+        if hasattr(self.model, "load_tokenizer"):
+            self.model.load_tokenizer(model_id, local_tokenizer_path)
+        else:
+            logger.error("The model wrapper does not support dynamic loading.")
+
     def fine_tune(self, data_path, fine_tune_config=None):
         """
         Fine-tune the agent's underlying model.
diff --git a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
index 7a38654d4..31a53c947 100644
--- a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
@@ -15,8 +15,9 @@ def main() -> None:
             {
                 "model_type": "huggingface",
                 "config_name": "my_custom_model",
-                "model_id": "google/gemma-2b-it",  # Or another generative model of your choice
-                # "local_model_path": # Specify your local model path
+                # "model_id": "google/gemma-2b-it",  # Or another generative model of your choice. Needed from loading from Hugging Face.
+                # "local_model_path":  # Specify your local model path
+                # "local_tokenizer_path":  # Specify your local tokenizer path
                 "max_length": 128,
                 "device": "cuda",
                 "data_path": "GAIR/lima", # Specify a Hugging Face data path if you wish to finetune the model from the start
@@ -39,7 +40,9 @@ def main() -> None:
         model_config_name="my_custom_model",  # Use your custom model config name here
     )
     
-    dialog_agent.load_model(model_id = "google/gemma-2b-it", local_model_path = None) #load gemma-2b-it from Hugging Face
+    dialog_agent.load_model(model_id = "google/gemma-2b-it", local_model_path = None) #load model gemma-2b-it from Hugging Face
+    dialog_agent.load_tokenizer(model_id = "google/gemma-2b-it", local_tokenizer_path = None) #load tokenizer for gemma-2b-it from Hugging Face
+
     # dialog_agent.fine_tune(data_path=  "GAIR/lima") #fine-tune loaded model with lima dataset with default hyperparameters
     
     #fine-tune loaded model with lima dataset with customized hyperparameters (`fine_tune_config` argument is optional. Defaults to None.)

From 5237356c116efaaea78aa019e2bcf29aebfeea55 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 25 Apr 2024 18:28:41 +0800
Subject: [PATCH 05/55] removed unnecessary info in README

---
 examples/load_finetune_huggingface_model/README.md | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/examples/load_finetune_huggingface_model/README.md b/examples/load_finetune_huggingface_model/README.md
index 400df461f..1373ed6c0 100644
--- a/examples/load_finetune_huggingface_model/README.md
+++ b/examples/load_finetune_huggingface_model/README.md
@@ -2,10 +2,6 @@
 
 This example demonstrates how to load and optionally fine-tune a Hugging Face model within a multi-agent conversation setup using AgentScope. The complete code is provided in `load_finetune_huggingface_model.py`.
 
-## Background
-
-In the context of AgentScope, agents are designed to mimic user and assistant roles in a conversation. This setup allows for the integration and testing of different models from the Hugging Face Hub, enhancing their capabilities through fine-tuning with custom datasets.
-
 ## Functionality Overview
 
 This example allows you to:

From a6918eb2b290fbf56fffb5c8d1c40f46822945b1 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 25 Apr 2024 22:31:22 +0800
Subject: [PATCH 06/55] resolved all issues flagged by `pre-commit run`

---
 .../huggingface_model.py                      | 498 ++++++++++++------
 .../load_finetune_huggingface_model.py        |  77 ++-
 2 files changed, 404 insertions(+), 171 deletions(-)

diff --git a/examples/load_finetune_huggingface_model/huggingface_model.py b/examples/load_finetune_huggingface_model/huggingface_model.py
index 8c48f6ec0..ceddb27b7 100644
--- a/examples/load_finetune_huggingface_model/huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/huggingface_model.py
@@ -1,66 +1,125 @@
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from agentscope.agents import DialogAgent
+# -*- coding: utf-8 -*-
+"""
+This module provides a HuggingFaceWrapper to manage
+and operate Hugging Face Transformers models, enabling loading,
+fine-tuning, and response generation. It includes the
+Finetune_DialogAgent class, which extends DialogAgent to
+enhance fine-tuning capabilities with custom hyperparameters.
+Key features include handling model and tokenizer operations,
+adapting to specialized datasets, and robust error management.
+
+Classes:
+- HuggingFaceWrapper: Manages Hugging Face models and tokenizers.
+- Finetune_DialogAgent: Extends DialogAgent for model fine-tuning.
+
+"""
+from typing import Optional, List, Dict, Any
+import os
 
-from agentscope.models import ModelWrapperBase, ModelResponse
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
 from loguru import logger
+from dotenv import load_dotenv
 
-import torch
-import os
-from dotenv import load_dotenv, find_dotenv
+from agentscope.agents import DialogAgent
+from agentscope.models import ModelWrapperBase, ModelResponse
 
-from typing import Optional
 
 class HuggingFaceWrapper(ModelWrapperBase):
     """Wrapper for a Hugging Face transformer model.
 
-    This class is responsible for loading and fine-tuning pre-trained models from the Hugging Face library.
+    This class is responsible for loading and fine-tuning
+    pre-trained models from the Hugging Face library.
     """
+
     model_type: str = "huggingface"  # Unique identifier for this model wrapper
 
-    def __init__(self, config_name, model_id = None, max_length=512, data_path = None, device = None, local_model_path = None, local_tokenizer_path = None, fine_tune_config=None, **kwargs):
+    def __init__(
+        self,
+        config_name: str,
+        model_id: Optional[str] = None,
+        max_length: int = 512,
+        data_path: Optional[str] = None,
+        device: Optional[torch.device] = None,
+        local_model_path: Optional[str] = None,
+        local_tokenizer_path: Optional[str] = None,
+        fine_tune_config: Optional[Dict[str, Any]] = None,
+        **kwargs: Any,
+    ) -> None:
         """Initializes the HuggingFaceWrapper with the given configuration.
 
         Arguments:
             config_name (str): Configuration name for model setup.
-            model_id (str): Identifier for the pre-trained model on Hugging Face.
-            max_length (int): Maximum sequence length for the model output per reply. Defaults to 512.
-            data_path (str, optional): Path to the dataset for fine-tuning the model.
-            device (torch.device, optional): Device to run the model on. Will default to GPU if available.
-            local_model_path (str, optional): Local file path to a pre-trained model.
-            fine_tune_config (dict, optional): Configuration for fine-tuning the model.
+            model_id (str): Identifier for the pre-trained model on
+                            Hugging Face.
+            max_length (int): Maximum sequence length for the
+                              model output per reply.
+                              Defaults to 512.
+            data_path (str, optional): Path to the dataset for
+                                       fine-tuning the model.
+            device (torch.device, optional): Device to run the model on.
+                                             Default to GPU if available.
+            local_model_path (str, optional): Local file path to a
+                                              pre-trained model.
+            fine_tune_config (dict, optional): Configuration for
+                                               fine-tuning the model.
             **kwargs: Additional keyword arguments.
         """
         super().__init__(config_name=config_name)
+        self.model = None
         self.max_length = max_length  # Set max_length as an attribute
         self.model_id = model_id
-        relative_path = os.path.join(os.path.dirname(__file__), "../load_finetune_huggingface_model/.env")
+        relative_path = os.path.join(
+            os.path.dirname(__file__),
+            "../load_finetune_huggingface_model/.env",
+        )
         dotenv_path = os.path.normpath(relative_path)
-        _ = load_dotenv(dotenv_path) # read local .env file
-        self.huggingface_token  = os.getenv('HUGGINGFACE_TOKEN')
+        _ = load_dotenv(dotenv_path)  # read local .env file
+        self.huggingface_token = os.getenv("HUGGINGFACE_TOKEN")
 
-        if device == None:
-            self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+        if device is None:
+            self.device = torch.device(
+                "cuda" if torch.cuda.is_available() else "cpu",
+            )
         else:
             self.device = device
-        
-        self.load_model(model_id, local_model_path = local_model_path)
-        self.load_tokenizer(model_id, local_tokenizer_path = local_tokenizer_path)
-        
-        if data_path != None:
-            self.model = self.fine_tune_training(self.model, self.tokenizer, data_path,  token = self.huggingface_token, fine_tune_config = fine_tune_config)
-                
-    def __call__(self, input, **kwargs) -> ModelResponse:
+
+        self.load_model(model_id, local_model_path=local_model_path)
+        self.load_tokenizer(
+            model_id,
+            local_tokenizer_path=local_tokenizer_path,
+        )
+
+        if data_path is not None:
+            self.model = self.fine_tune_training(
+                self.model,
+                self.tokenizer,
+                data_path,
+                token=self.huggingface_token,
+                fine_tune_config=fine_tune_config,
+            )
+
+    def __call__(
+        self,
+        inputs: List[Dict[str, Any]],
+        **kwargs: Any,
+    ) -> ModelResponse:
         """Process the input data to generate a response from the model.
 
-        This method tokenizes the input text, generates a response using the model,
+        This method tokenizes the input text, generates
+        a response using the model,
         and then decodes the generated tokens into a string.
 
         Arguments:
-            input (list): A list of dictionaries where each dictionary contains 'name', 'role' and 'content' keys and their respective values.
-            **kwargs: Additional keyword arguments for the model's generate function.
+            input (list): A list of dictionaries where each dictionary contains
+                          'name', 'role' and 'content' keys
+                          and their respective values.
+            **kwargs: Additional keyword arguments for the
+                      model's generate function.
 
         Returns:
-            ModelResponse: An object containing the generated text and raw model output.
+            ModelResponse: An object containing the generated
+                           text and raw model output.
 
         Raises:
             Exception: If an error occurs during text generation.
@@ -68,123 +127,229 @@ def __call__(self, input, **kwargs) -> ModelResponse:
 
         try:
             # Tokenize the input text
-            concatenated_input  = "\n".join([f"{d.get('name', 'System')}: {d['content']}" for d in input])
-            input_ids = self.tokenizer.encode(f"{concatenated_input}\nAssistent: ", return_tensors='pt')
+            concatenated_input = "\n".join(
+                [f"{d.get('name', 'System')}: {d['content']}" for d in inputs],
+            )
+            input_ids = self.tokenizer.encode(
+                f"{concatenated_input}\nAssistent: ",
+                return_tensors="pt",
+            )
             # Generate response using the model
-            outputs = self.model.generate(input_ids.to(self.device), max_new_tokens = self.max_length, **kwargs)
+            outputs = self.model.generate(
+                input_ids.to(self.device),
+                max_new_tokens=self.max_length,
+                **kwargs,
+            )
             # Decode the generated tokens to a string
-            generated_text = self.tokenizer.decode(outputs[0][input_ids.shape[1]:], skip_special_tokens=True)
-            
-            return ModelResponse(text=generated_text, raw = outputs)
+            generated_text = self.tokenizer.decode(
+                outputs[0][input_ids.shape[1]:],
+                skip_special_tokens=True,
+            )
+            return ModelResponse(text=generated_text, raw=outputs)
         except Exception as e:
             logger.error(f"Generation error: {e}")
             raise
 
-    def load_model(self, model_id, local_model_path = None):
+    def format(self, data: Any) -> Any:
         """
-        Load a new model for the agent from a local path and update the agent's model.
-        
+        Pass-through for data formatting. Assume
+        data is already in the correct format.
+
+        Arguments:
+            data (Any): Data to be formatted.
+
+        Returns:
+            Any: The input data unchanged.
+        """
+        return data
+
+    def load_model(
+        self,
+        model_id: Optional[str] = None,
+        local_model_path: Optional[str] = None,
+    ) -> None:
+        """
+        Load a new model for the agent from
+        a local path and update the agent's model.
+
         Arguments:
             local_model_path (str): The file path to the model to be loaded.
             model_id (str): An identifier for the model on Huggingface.
-        
+
         Raises:
-            Exception: If the model cannot be loaded from the given path or identifier. Possible reasons include file not found, incorrect model ID, or network issues while fetching the model.
+            Exception: If the model cannot be loaded from the given
+                       path or identifier.
+                       Possible reasons include file not found,
+                       incorrect model ID,
+                       or network issues while fetching the model.
         """
 
         try:
-            if local_model_path == None:
-                self.model = AutoModelForCausalLM.from_pretrained(model_id, token = self.huggingface_token, device_map="auto",)
+            if local_model_path is None:
+                self.model = AutoModelForCausalLM.from_pretrained(
+                    model_id,
+                    token=self.huggingface_token,
+                    device_map="auto",
+                )
                 print("new model")
             else:
                 self.model = AutoModelForCausalLM.from_pretrained(
-                    local_model_path, local_files_only=True,
-                    device_map="auto"
+                    local_model_path,
+                    local_files_only=True,
+                    device_map="auto",
                 )
                 print("local model")
-            
-            #log the successful model loading
-            logger.info(f"Successfully loaded new model '{model_id}' from '{local_model_path}'")
+
+            # log the successful model loading
+            info_msg = (
+                f"Successfully loaded new model '{model_id}' from "
+                f"'{local_model_path}'"
+            )
+            logger.info(info_msg)
+
         except Exception as e:
-            # Handle exceptions during model loading, such as file not found or load errors
-            logger.error(f"Failed to load model '{model_id}' from '{local_model_path}': {e}")
-            raise  # Or handle error appropriately
-    
-    def load_tokenizer(self, model_id, local_tokenizer_path = None):
+            # Handle exceptions during model loading,
+            # such as file not found or load errors
+            error_msg = (
+                f"Failed to load model '{model_id}' "
+                f"from '{local_model_path}': {e}"
+            )
+
+            logger.error(error_msg)
+
+            raise
+
+    def load_tokenizer(
+        self,
+        model_id: Optional[str] = None,
+        local_tokenizer_path: Optional[str] = None,
+    ) -> None:
         """
         Load the tokenizer from a local path.
-        
+
         Arguments:
-            local_tokenizer_path (str): The file path to the tokenizer to be loaded.
+            local_tokenizer_path (str): The file path to the
+                                        tokenizer to be loaded.
             model_id (str): An identifier for the model on Huggingface.
-        
+
         Raises:
-            Exception: If the tokenizer cannot be loaded from the given path or identifier. Possible reasons include file not found, incorrect model ID, or network issues while fetching the tokenizer.
+            Exception: If the tokenizer cannot be loaded from the
+            given path or identifier. Possible reasons include file not found,
+            incorrect model ID, or network issues while fetching the tokenizer.
         """
 
         try:
-            if local_tokenizer_path == None:
-                self.tokenizer = AutoTokenizer.from_pretrained(model_id, token = self.huggingface_token)
+            if local_tokenizer_path is None:
+                self.tokenizer = AutoTokenizer.from_pretrained(
+                    model_id,
+                    token=self.huggingface_token,
+                )
                 print("new tokenizer")
             else:
-                self.tokenizer = AutoTokenizer.from_pretrained(local_tokenizer_path)
+                self.tokenizer = AutoTokenizer.from_pretrained(
+                    local_tokenizer_path,
+                )
                 print("local tokenizer")
-            
-            #log the successful tokenizer loading
-            logger.info(f"Successfully loaded new tokenizer for model '{model_id}' from '{local_tokenizer_path}'")
+
+            # log the successful tokenizer loading
+            logger.info(
+                f"Successfully loaded new tokenizer for model '{model_id}' "
+                f"from '{local_tokenizer_path}'",
+            )
+
         except Exception as e:
-            # Handle exceptions during model loading, such as file not found or load errors
-            logger.error(f"Failed to load tokenizer for model '{model_id}' from '{local_tokenizer_path}': {e}")
-            raise  # Or handle error appropriately
+            # Handle exceptions during model loading,
+            # such as file not found or load errors
+            error_message = (
+                f"Failed to load tokenizer for model '{model_id}' from "
+                f"'{local_tokenizer_path}': {e}"
+            )
+            logger.error(error_message)
+
+            raise
 
-    def fine_tune(self, data_path, fine_tune_config=None):
+    def fine_tune(
+        self,
+        data_path: Optional[str] = None,
+        fine_tune_config: Optional[Dict[str, Any]] = None,
+    ) -> None:
         """
         Fine-tune the agent's model using data from the specified path.
-        
+
         Arguments:
-            data_path (str): The file path to the training data from Hugging Face.
-            
+            data_path (str): The file path to the training
+                             data from Hugging Face.
+
         Raises:
-            Exception: If the fine-tuning process fails. This could be due to issues with the data path, configuration parameters, or internal errors during the training process.
+            Exception: If the fine-tuning process fails. This could be
+            due to issues with the data path, configuration parameters,
+            or internal errors during the training process.
         """
         try:
-            self.model = self.fine_tune_training(self.model, self.tokenizer, data_path, token = self.huggingface_token, fine_tune_config = fine_tune_config)
-            
-            logger.info(f"Successfully fine-tuned model with data from '{data_path}'")
+            self.model = self.fine_tune_training(
+                self.model,
+                self.tokenizer,
+                data_path,
+                token=self.huggingface_token,
+                fine_tune_config=fine_tune_config,
+            )
+
+            logger.info(
+                f"Successfully fine-tuned model with data from '{data_path}'",
+            )
         except Exception as e:
-            logger.error(f"Failed to fine-tune model with data from '{data_path}': {e}")
-            raise  # Or handle the error appropriately
-
+            logger.error(
+                f"Failed to fine-tune model with data from '{data_path}': {e}",
+            )
+            raise
 
-    def fine_tune_training(self, model, tokenizer, data_path, token, fine_tune_config=None):
-        
+    def fine_tune_training(
+        self,
+        model: AutoModelForCausalLM,
+        tokenizer: AutoTokenizer,
+        data_path: Optional[str] = None,
+        token: Optional[str] = None,
+        fine_tune_config: Optional[Dict[str, Any]] = None,
+    ) -> AutoModelForCausalLM:
         """
-        The actual method that handles training and fine-tuning the model on the dataset specified by the data_path using a given tokenizer.
+        The actual method that handles training and fine-tuning
+        the model on the dataset specified by
+        the data_path using a given tokenizer.
 
         Arguments:
-            model (AutoModelForCausalLM): The pre-trained causal language model from Hugging Face's transformers.
-            tokenizer (AutoTokenizer): The tokenizer corresponding to the pre-trained model.
-            data_path (str): The file path or dataset identifier to load the dataset from Hugging Face.
+            model (AutoModelForCausalLM): The pre-trained causal language model
+                                          from Hugging Face's transformers.
+            tokenizer (AutoTokenizer): The tokenizer corresponding to
+                                       the pre-trained model.
+            data_path (str): The file path or dataset identifier to load
+                             the dataset from Hugging Face.
             token (str): The authentication token for Hugging Face.
-            fine_tune_config (dict, optional): Configuration options for fine-tuning the model, including LoRA and training arguments.
+            fine_tune_config (dict, optional): Configuration options for
+                                               fine-tuning the model,
+                                               including LoRA and training
+                                               arguments.
 
         Returns:
             AutoModelForCausalLM: The fine-tuned language model.
 
         Raises:
-            Exception: Raises an exception if the dataset loading or fine-tuning process fails.
+            Exception: Raises an exception if the dataset
+                       loading or fine-tuning process fails.
 
         Note:
-            This method updates the model in place and also logs the fine-tuning process.
-            It utilizes the LoRA configuration and custom training arguments to adapt the pre-trained model to the specific dataset.
-            The training log and trained model are saved in the same directory with the specific timestamp at saving time as part of the log/model fodler name. 
+            This method updates the model in place and also logs
+            the fine-tuning process.
+            It utilizes the LoRA configuration and custom training arguments
+            to adapt the pre-trained model to the specific dataset.
+            The training log and trained model are saved in the same
+            directory with the specific timestamp at saving time
+            as part of the log/model fodler name.
         """
         from datasets import load_dataset
         from datetime import datetime
-        import os
         import json
 
-        dataset = load_dataset(data_path, token = token)
+        dataset = load_dataset(data_path, token=token)
 
         from peft import LoraConfig
 
@@ -193,13 +358,12 @@ def fine_tune_training(self, model, tokenizer, data_path, token, fine_tune_confi
             "lora_alpha": 32,
             "lora_dropout": 0.05,
             "bias": "none",
-            "task_type": "CAUSAL_LM"
+            "task_type": "CAUSAL_LM",
         }
 
         if fine_tune_config is not None:
-            if fine_tune_config['lora_config'] is not None:
-                lora_config_default.update(fine_tune_config['lora_config'])
-
+            if fine_tune_config["lora_config"] is not None:
+                lora_config_default.update(fine_tune_config["lora_config"])
 
         training_defaults = {
             "per_device_train_batch_size": 1,
@@ -215,27 +379,33 @@ def fine_tune_training(self, model, tokenizer, data_path, token, fine_tune_confi
         }
 
         if fine_tune_config is not None:
-            if fine_tune_config['training_args'] is not None:
-                training_defaults.update(fine_tune_config['training_args'])
+            if fine_tune_config["training_args"] is not None:
+                training_defaults.update(fine_tune_config["training_args"])
 
         from peft import get_peft_model
 
-        
         lora_config = LoraConfig(**lora_config_default)
         model = get_peft_model(model, lora_config)
 
         from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
         import transformers
 
-        def formatting_prompts_func(example):
+        def formatting_prompts_func(
+            example: Dict[str, List[List[str]]],
+        ) -> List[str]:
             output_texts = []
-            for i in range(len(example['conversations'])):
-                text = f"### Question: {example['conversations'][i][0]}\n ### Answer: {example['conversations'][i][1]}"
+            for i in range(len(example["conversations"])):
+                question = f"### Question: {example['conversations'][i][0]}"
+                answer = f"### Answer: {example['conversations'][i][1]}"
+                text = f"{question}\n {answer}"
                 output_texts.append(text)
             return output_texts
 
         response_template = " ### Answer:"
-        collator = DataCollatorForCompletionOnlyLM(response_template, tokenizer=tokenizer)
+        collator = DataCollatorForCompletionOnlyLM(
+            response_template,
+            tokenizer=tokenizer,
+        )
 
         trainer_args = transformers.TrainingArguments(**training_defaults)
 
@@ -253,93 +423,122 @@ def formatting_prompts_func(example):
 
         trainer.train()
 
-        
         now = datetime.now()
-        time_string = now.strftime('%Y-%m-%d_%H-%M-%S')
+        time_string = now.strftime("%Y-%m-%d_%H-%M-%S")
 
         # Specify the filename
-        log_name = f"{model.config._name_or_path.split('/')[-1]}_{time_string}_log_history.json"
+        log_name_temp = model.config.name_or_path.split("/")[-1]
+        log_name = f"{log_name_temp}_{time_string}_log_history.json"
         log_path = os.path.join(os.path.dirname(__file__), log_name)
 
-        # Writing JSON data
-        with open(log_path, 'w') as f:
+        # log training history
+        with open(log_path, "w", encoding="utf-8") as f:
             json.dump(trainer.state.log_history, f)
 
-        
-
-        # os.makedirs(os.path.dirname(model_path), exist_ok=True)
-        # Check if directory exists
-        # if not os.path.exists(model_path):
-        #     # If not, create the directory
-        #     os.makedirs(model_path)
-
-        #save model
-        model_name = f"sft_{model.config._name_or_path.split('/')[-1]}_{time_string}"
+        # save model
+        model_name = (
+            f"sft_{model.config.name_or_path.split('/')[-1]}_{time_string}"
+        )
         model_path = os.path.join(os.path.dirname(__file__), model_name)
         trainer.save_model(model_path)
 
-        #save tokenizer
-        tokenizer_name = f"sft_{model.config._name_or_path.split('/')[-1]}_tokenizer_{time_string}"
-        tokenizer_path = os.path.join(os.path.dirname(__file__), tokenizer_name)
+        # save tokenizer
+        tokenizer_name_temp = model.config.name_or_path.split("/")[-1]
+        tokenizer_name = f"sft_{tokenizer_name_temp}_tokenizer_{time_string}"
+        tokenizer_path = os.path.join(
+            os.path.dirname(__file__),
+            tokenizer_name,
+        )
         tokenizer.save_pretrained(tokenizer_path)
 
-        
         return model
 
 
 class Finetune_DialogAgent(DialogAgent):
     """
-    A dialog agent capable of fine-tuning its underlying model based on provided data.
+    A dialog agent capable of fine-tuning its
+    underlying model based on provided data.
 
-    Inherits from DialogAgent and adds functionality for fine-tuning with custom hyperparameters.
+    Inherits from DialogAgent and adds functionality for
+    fine-tuning with custom hyperparameters.
     """
 
-    def __init__(self, name: str, sys_prompt: str, model_config_name: str, use_memory: bool = True, memory_config: Optional[dict] = None):
+    def __init__(
+        self,
+        name: str,
+        sys_prompt: str,
+        model_config_name: str,
+        use_memory: bool = True,
+        memory_config: Optional[dict] = None,
+    ):
         """
         Initializes a new Finetune_DialogAgent with specified configuration.
 
         Arguments:
             name (str): Name of the agent.
             sys_prompt (str): System prompt or description of the agent's role.
-            model_config_name (str): The configuration name for the underlying model.
-            use_memory (bool, optional): Indicates whether to utilize memory features. Defaults to True.
-            memory_config (dict, optional): Configuration for memory functionalities if `use_memory` is True.
-        
+            model_config_name (str): The configuration name for
+                                     the underlying model.
+            use_memory (bool, optional): Indicates whether to utilize
+                                         memory features. Defaults to True.
+            memory_config (dict, optional): Configuration for memory
+                                            functionalities if
+                                            `use_memory` is True.
+
         Note:
             Refer to `class DialogAgent(AgentBase)` for more information.
         """
-        
-        super().__init__(name, sys_prompt, model_config_name, use_memory, memory_config)
-        
-        
 
-    def load_model(self, model_id = None, local_model_path=None):
+        super().__init__(
+            name,
+            sys_prompt,
+            model_config_name,
+            use_memory,
+            memory_config,
+        )
+
+    def load_model(
+        self,
+        model_id: Optional[str] = None,
+        local_model_path: Optional[str] = None,
+    ) -> None:
         """
         Load a new model into the agent.
 
         Arguments:
-            model_id (str): The Hugging Face model ID or a custom identifier. Needed if loading model from Hugging Face.
+            model_id (str): The Hugging Face model ID or a custom identifier.
+                            Needed if loading model from Hugging Face.
             local_model_path (str, optional): Path to a locally saved model.
-        
+
         Raises:
-            Exception: If the model loading process fails or if the model wrapper does not support dynamic loading.
+            Exception: If the model loading process fails or if the
+                       model wrapper does not support dynamic loading.
         """
 
         if hasattr(self.model, "load_model"):
             self.model.load_model(model_id, local_model_path)
         else:
-            logger.error("The model wrapper does not support dynamic model loading.")
-
-    def load_tokenizer(self, model_id = None, local_tokenizer_path=None):
+            logger.error(
+                "The model wrapper does not support dynamic model loading.",
+            )
+
+    def load_tokenizer(
+        self,
+        model_id: Optional[str] = None,
+        local_tokenizer_path: Optional[str] = None,
+    ) -> None:
         """
         Load a new tokenizer for the agent.
 
         Arguments:
-            model_id (str): The Hugging Face model ID or a custom identifier. Needed if loading tokenizer from Hugging Face.
-            local_tokenizer_path (str, optional): Path to a locally saved tokenizer.
-        
+            model_id (str): The Hugging Face model ID or a custom identifier.
+                            Needed if loading tokenizer from Hugging Face.
+            local_tokenizer_path (str, optional): Path to a locally saved
+                                                  tokenizer.
+
         Raises:
-            Exception: If the model tokenizer process fails or if the model wrapper does not support dynamic loading.
+            Exception: If the model tokenizer process fails or if the
+                       model wrapper does not support dynamic loading.
         """
 
         if hasattr(self.model, "load_tokenizer"):
@@ -347,15 +546,20 @@ def load_tokenizer(self, model_id = None, local_tokenizer_path=None):
         else:
             logger.error("The model wrapper does not support dynamic loading.")
 
-    def fine_tune(self, data_path, fine_tune_config=None):
+    def fine_tune(
+        self,
+        data_path: Optional[str] = None,
+        fine_tune_config: Optional[Dict[str, Any]] = None,
+    ) -> None:
         """
         Fine-tune the agent's underlying model.
 
         Arguments:
             data_path (str): The path to the training data.
-        
+
         Raises:
-            Exception: If fine-tuning fails or if the model wrapper does not support fine-tuning.
+            Exception: If fine-tuning fails or if the
+                       model wrapper does not support fine-tuning.
         """
 
         if hasattr(self.model, "fine_tune"):
diff --git a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
index 31a53c947..b16f51f26 100644
--- a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
@@ -1,34 +1,50 @@
+# -*- coding: utf-8 -*-
+"""
+This script sets up a conversational agent using
+AgentScope with a Hugging Face model.
+It includes initializing a Finetune_DialogAgent,
+loading and fine-tuning a pre-trained model,
+and conducting a dialogue via a sequential pipeline.
+The conversation continues until the user exits.
+Features include model and tokenizer loading,
+and fine-tuning on the GAIR/lima dataset with adjustable parameters.
+"""
+from huggingface_model import Finetune_DialogAgent
+
 import agentscope
 from agentscope.agents.user_agent import UserAgent
 from agentscope.pipelines.functional import sequentialpipeline
-from huggingface_model import Finetune_DialogAgent
 
 
 def main() -> None:
     """A basic conversation demo with a custom model"""
 
     # Initialize AgentScope with your custom model configuration
-    
 
     agentscope.init(
         model_configs=[
             {
                 "model_type": "huggingface",
                 "config_name": "my_custom_model",
-                # "model_id": "google/gemma-2b-it",  # Or another generative model of your choice. Needed from loading from Hugging Face.
+                # Or another generative model of your choice.
+                # Needed from loading from Hugging Face.
+                "model_id": "google/gemma-2b-it",
                 # "local_model_path":  # Specify your local model path
                 # "local_tokenizer_path":  # Specify your local tokenizer path
                 "max_length": 128,
                 "device": "cuda",
-                "data_path": "GAIR/lima", # Specify a Hugging Face data path if you wish to finetune the model from the start
-
-                # fine_tune_config (Optional): Configuration for fine-tuning the model. This dictionary
-                # can include hyperparameters and other training options that
-                # will be passed to the fine-tuning method. Defaults to None.
-                "fine_tune_config":{
-                "lora_config": {"r": 20, "lora_alpha": 40},
-                "training_args": {"max_steps": 20, "logging_steps": 2}
-                }
+                # Specify a Hugging Face data path if you
+                # wish to finetune the model from the start
+                "data_path": "GAIR/lima",
+                # fine_tune_config (Optional): Configuration for
+                # fine-tuning the model.
+                # This dictionary can include hyperparameters and other
+                # training options that will be passed to the
+                # fine-tuning method. Defaults to None.
+                "fine_tune_config": {
+                    "lora_config": {"r": 20, "lora_alpha": 40},
+                    "training_args": {"max_steps": 20, "logging_steps": 2},
+                },
             },
         ],
     )
@@ -37,19 +53,31 @@ def main() -> None:
     dialog_agent = Finetune_DialogAgent(
         name="Assistant",
         sys_prompt="You're a helpful assistant.",
-        model_config_name="my_custom_model",  # Use your custom model config name here
+        # Use your custom model config name here
+        model_config_name="my_custom_model",
+    )
+
+    dialog_agent.load_model(
+        model_id="google/gemma-2b-it",
+        local_model_path=None,
+    )  # load model gemma-2b-it from Hugging Face
+    dialog_agent.load_tokenizer(
+        model_id="google/gemma-2b-it",
+        local_tokenizer_path=None,
+    )  # load tokenizer for gemma-2b-it from Hugging Face
+
+    # fine-tune loaded model with lima dataset with default hyperparameters
+    # dialog_agent.fine_tune(data_path=  "GAIR/lima")
+
+    # fine-tune loaded model with lima dataset with customized hyperparameters
+    # (`fine_tune_config` argument is optional. Defaults to None.)
+    dialog_agent.fine_tune(
+        "GAIR/lima",
+        fine_tune_config={
+            "lora_config": {"r": 24, "lora_alpha": 48},
+            "training_args": {"max_steps": 30, "logging_steps": 3},
+        },
     )
-    
-    dialog_agent.load_model(model_id = "google/gemma-2b-it", local_model_path = None) #load model gemma-2b-it from Hugging Face
-    dialog_agent.load_tokenizer(model_id = "google/gemma-2b-it", local_tokenizer_path = None) #load tokenizer for gemma-2b-it from Hugging Face
-
-    # dialog_agent.fine_tune(data_path=  "GAIR/lima") #fine-tune loaded model with lima dataset with default hyperparameters
-    
-    #fine-tune loaded model with lima dataset with customized hyperparameters (`fine_tune_config` argument is optional. Defaults to None.)
-    dialog_agent.fine_tune("GAIR/lima", fine_tune_config ={
-    "lora_config": {"r": 24, "lora_alpha": 48},
-    "training_args": {"max_steps": 30, "logging_steps": 3}
-    })
 
     user_agent = UserAgent()
 
@@ -58,5 +86,6 @@ def main() -> None:
     while x is None or x.content != "exit":
         x = sequentialpipeline([dialog_agent, user_agent], x)
 
+
 if __name__ == "__main__":
     main()

From b4f4f40ae56896b73b735d84d6d0c65137ed8517 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 25 Apr 2024 22:40:13 +0800
Subject: [PATCH 07/55] further removed info irrelevant to model loading and
 finetuning

---
 examples/load_finetune_huggingface_model/README.md | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)

diff --git a/examples/load_finetune_huggingface_model/README.md b/examples/load_finetune_huggingface_model/README.md
index 1373ed6c0..bdb6172d6 100644
--- a/examples/load_finetune_huggingface_model/README.md
+++ b/examples/load_finetune_huggingface_model/README.md
@@ -4,15 +4,7 @@ This example demonstrates how to load and optionally fine-tune a Hugging Face mo
 
 ## Functionality Overview
 
-This example allows you to:
-
-- Set up a user agent and an assistant agent for interactive conversations.
-- Modify the `sys_prompt` to customize the assistant agent's role.
-- Terminate the conversation by entering "exit".
-
-## Advanced Features
-
-Beyond basic conversation setup, the example introduces advanced functionalities:
+Compared to basic conversation setup, this example introduces model loading and fine-tuning features:
 
 - Use `dialog_agent.load_model(model_id, local_model_path)` to load a model either from the Hugging Face Model Hub or a local directory.
 - Apply `dialog_agent.fine_tune(data_path)` to fine-tune the model based on your dataset.

From e33b3ded915b2c75375a27c1de27ea27c2cd1fef Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Fri, 26 Apr 2024 12:43:04 +0800
Subject: [PATCH 08/55] Update huggingface_model.py

fixed issue related to `format` method
---
 .../huggingface_model.py                      | 96 ++++++++++++++-----
 1 file changed, 70 insertions(+), 26 deletions(-)

diff --git a/examples/load_finetune_huggingface_model/huggingface_model.py b/examples/load_finetune_huggingface_model/huggingface_model.py
index ceddb27b7..1a95ce7eb 100644
--- a/examples/load_finetune_huggingface_model/huggingface_model.py
+++ b/examples/load_finetune_huggingface_model/huggingface_model.py
@@ -13,7 +13,7 @@
 - Finetune_DialogAgent: Extends DialogAgent for model fine-tuning.
 
 """
-from typing import Optional, List, Dict, Any
+from typing import Sequence, Any, Union, List, Optional, Dict
 import os
 
 import torch
@@ -23,6 +23,8 @@
 
 from agentscope.agents import DialogAgent
 from agentscope.models import ModelWrapperBase, ModelResponse
+from agentscope.message import MessageBase
+from agentscope.utils.tools import _convert_to_str
 
 
 class HuggingFaceWrapper(ModelWrapperBase):
@@ -127,11 +129,11 @@ def __call__(
 
         try:
             # Tokenize the input text
-            concatenated_input = "\n".join(
-                [f"{d.get('name', 'System')}: {d['content']}" for d in inputs],
+            concatenated_input = "\n ".join(
+                [f"{d.get('role')}: {d['content']}" for d in inputs],
             )
             input_ids = self.tokenizer.encode(
-                f"{concatenated_input}\nAssistent: ",
+                f"{concatenated_input}\n assistent: ",
                 return_tensors="pt",
             )
             # Generate response using the model
@@ -150,18 +152,56 @@ def __call__(
             logger.error(f"Generation error: {e}")
             raise
 
-    def format(self, data: Any) -> Any:
-        """
-        Pass-through for data formatting. Assume
-        data is already in the correct format.
+    def format(
+        self,
+        *args: Union[MessageBase, Sequence[MessageBase]],
+    ) -> List[dict]:
+        """A basic strategy to format the input into the required format of
+        Hugging Face models.
 
-        Arguments:
-            data (Any): Data to be formatted.
+        Args:
+            args (`Union[MessageBase, Sequence[MessageBase]]`):
+                The input arguments to be formatted, where each argument
+                should be a `Msg` object, or a list of `Msg` objects.
+                In distribution, placeholder is also allowed.
 
         Returns:
-            Any: The input data unchanged.
+            `List[dict]`:
+                The formatted messages.
         """
-        return data
+        huggingface_msgs = []
+        for msg in args:
+            if msg is None:
+                continue
+            if isinstance(msg, MessageBase):
+                # content shouldn't be empty string
+                if msg.content == "":
+                    logger.warning(
+                        "The content field cannot be "
+                        "empty string. To avoid error, the empty string is "
+                        "replaced by a blank space automatically, but the "
+                        "model may not work as expected.",
+                    )
+                    msg.content = " "
+
+                huggingface_msg = {
+                    "role": msg.role,
+                    "content": _convert_to_str(msg.content),
+                }
+
+                # image url
+                if msg.url is not None:
+                    huggingface_msg["images"] = [msg.url]
+
+                huggingface_msgs.append(huggingface_msg)
+            elif isinstance(msg, list):
+                huggingface_msgs.extend(self.format(*msg))
+            else:
+                raise TypeError(
+                    f"Invalid message type: {type(msg)}, `Msg` is expected.",
+                )
+
+        return huggingface_msgs
 
     def load_model(
         self,
@@ -191,20 +231,22 @@ def load_model(
                     token=self.huggingface_token,
                     device_map="auto",
                 )
-                print("new model")
+                info_msg = (
+                    f"Successfully loaded new model '{model_id}' from "
+                    f"Hugging Face"
+                )
             else:
                 self.model = AutoModelForCausalLM.from_pretrained(
                     local_model_path,
                     local_files_only=True,
                     device_map="auto",
                 )
-                print("local model")
+                info_msg = (
+                    f"Successfully loaded new model '{model_id}' from "
+                    f"'{local_model_path}'"
+                )
 
             # log the successful model loading
-            info_msg = (
-                f"Successfully loaded new model '{model_id}' from "
-                f"'{local_model_path}'"
-            )
             logger.info(info_msg)
 
         except Exception as e:
@@ -244,18 +286,20 @@ def load_tokenizer(
                     model_id,
                     token=self.huggingface_token,
                 )
-                print("new tokenizer")
+                # log the successful tokenizer loading
+                logger.info(
+                    f"Successfully loaded new tokenizer for model "
+                    f"'{model_id}' from Hugging Face",
+                )
             else:
                 self.tokenizer = AutoTokenizer.from_pretrained(
                     local_tokenizer_path,
                 )
-                print("local tokenizer")
-
-            # log the successful tokenizer loading
-            logger.info(
-                f"Successfully loaded new tokenizer for model '{model_id}' "
-                f"from '{local_tokenizer_path}'",
-            )
+                # log the successful tokenizer loading
+                logger.info(
+                    f"Successfully loaded new tokenizer for model "
+                    f"'{model_id}' from '{local_tokenizer_path}'",
+                )
 
         except Exception as e:
             # Handle exceptions during model loading,

From 80238205d6fb9e9d63e61f8b62026cfb97d34123 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 2 May 2024 11:49:19 +0800
Subject: [PATCH 09/55] updated according to suggestions given

---
 .../configs/model_configs.json                |  18 ++
 ...rsation_with_agent_with_finetuned_model.py |  98 +++++++++
 .../finetune_dialogagent.py                   | 132 ++++++++++++
 .../huggingface_model.py                      | 188 +++++-------------
 .../load_finetune_huggingface_model/README.md |  42 ----
 .../load_finetune_huggingface_model.py        |  91 ---------
 6 files changed, 298 insertions(+), 271 deletions(-)
 create mode 100644 examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
 create mode 100644 examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
 create mode 100644 examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
 rename examples/{load_finetune_huggingface_model => conversation_with_agent_with_finetuned_model}/huggingface_model.py (77%)
 delete mode 100644 examples/load_finetune_huggingface_model/README.md
 delete mode 100644 examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py

diff --git a/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json b/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
new file mode 100644
index 000000000..3b59c8318
--- /dev/null
+++ b/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
@@ -0,0 +1,18 @@
+[
+    {
+        "model_type": "huggingface",
+        "config_name": "my_custom_model",
+
+        "model_id": "openlm-research/open_llama_3b_v2",
+
+        "max_length": 128,
+        "device": "cuda",
+
+        "data_path": "databricks/databricks-dolly-15k",
+
+        "fine_tune_config": {
+            "lora_config": {"r": 20, "lora_alpha": 40},
+            "training_args": {"max_steps": 1000, "logging_steps": 1}
+        }
+    }
+]
\ No newline at end of file
diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
new file mode 100644
index 000000000..989e8525f
--- /dev/null
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -0,0 +1,98 @@
+# -*- coding: utf-8 -*-
+"""
+This script sets up a conversational agent using
+AgentScope with a Hugging Face model.
+It includes initializing a Finetune_DialogAgent,
+loading and fine-tuning a pre-trained model,
+and conducting a dialogue via a sequential pipeline.
+The conversation continues until the user exits.
+Features include model and tokenizer loading,
+and fine-tuning on the databricks-dolly-15k dataset with adjustable parameters.
+"""
+from finetune_dialogagent import Finetune_DialogAgent
+
+import agentscope
+from agentscope.agents.user_agent import UserAgent
+from agentscope.pipelines.functional import sequentialpipeline
+
+
+def main() -> None:
+    """A basic conversation demo with a custom model"""
+
+    # Initialize AgentScope with your custom model configuration
+
+    # agentscope.init(
+    #     model_configs=[
+    #         {
+    #             "model_type": "huggingface",
+    #             "config_name": "my_custom_model",
+    #             # Or another generative model of your choice.
+    #             # Needed from loading from Hugging Face.
+    #             "model_id": "openlm-research/open_llama_3b_v2",
+    #             # "local_model_path":  # Specify your local model path
+    #             # "local_tokenizer_path":  # Specify your local tokenizer path
+    #             "max_length": 128,
+    #             "device": "cuda",
+    #             # Specify a Hugging Face data path if you
+    #             # wish to finetune the model from the start
+    #             "data_path": "databricks/databricks-dolly-15k",
+    #             # fine_tune_config (Optional): Configuration for
+    #             # fine-tuning the model.
+    #             # This dictionary can include hyperparameters and other
+    #             # training options that will be passed to the
+    #             # fine-tuning method. Defaults to None.
+    #             # `lora_config` and `training_args` follow 
+    #             # the standard lora and sfttrainer fields.
+    #             "fine_tune_config": {
+    #                 "lora_config": {"r": 20, "lora_alpha": 40},
+    #                 "training_args": {"max_steps": 1000, "logging_steps": 1},
+    #             },
+    #         },
+    #     ],
+    # )
+
+    # alternatively can load `model_configs` from json file
+    agentscope.init(
+    model_configs="./configs/model_configs.json",
+    )
+
+    # Init agents with the custom model
+    dialog_agent = Finetune_DialogAgent(
+        name="Assistant",
+        sys_prompt="You're a helpful assistant.",
+        # Use your custom model config name here
+        model_config_name="my_custom_model",
+    )
+
+    dialog_agent.load_model(
+        model_id="openlm-research/open_llama_3b_v2",
+        local_model_path=None,
+    )  # load model gemma-2b-it from Hugging Face
+    dialog_agent.load_tokenizer(
+        model_id="openlm-research/open_llama_3b_v2",
+        local_tokenizer_path=None,
+    )  # load tokenizer for gemma-2b-it from Hugging Face
+
+    # fine-tune loaded model with databricks-dolly-15k dataset with default hyperparameters
+    dialog_agent.fine_tune(data_path="databricks/databricks-dolly-15k")
+
+    # fine-tune loaded model with databricks-dolly-15k dataset with customized hyperparameters
+    # (`fine_tune_config` argument is optional. Defaults to None.)
+    dialog_agent.fine_tune(
+        "databricks/databricks-dolly-15k",
+        fine_tune_config={
+            "lora_config": {"r": 24, "lora_alpha": 48},
+            "training_args": {"max_steps": 30, "logging_steps": 3},
+        },
+    )
+
+    user_agent = UserAgent()
+
+    # Start the conversation between user and assistant
+    x = None
+    while x is None or x.content != "exit":
+        x = sequentialpipeline([dialog_agent, user_agent], x)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py b/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
new file mode 100644
index 000000000..1059cc99e
--- /dev/null
+++ b/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
@@ -0,0 +1,132 @@
+# -*- coding: utf-8 -*-
+"""
+This module provides the Finetune_DialogAgent class, 
+which extends DialogAgent to enhance fine-tuning 
+capabilities with custom hyperparameters.
+"""
+from typing import Sequence, Any, Union, List, Optional, Dict
+import os
+
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from loguru import logger
+from dotenv import load_dotenv
+
+from agentscope.agents import DialogAgent
+from agentscope.models import ModelWrapperBase, ModelResponse
+from agentscope.message import MessageBase
+from agentscope.utils.tools import _convert_to_str
+
+class Finetune_DialogAgent(DialogAgent):
+    """
+    A dialog agent capable of fine-tuning its
+    underlying model based on provided data.
+
+    Inherits from DialogAgent and adds functionality for
+    fine-tuning with custom hyperparameters.
+    """
+
+    def __init__(
+        self,
+        name: str,
+        sys_prompt: str,
+        model_config_name: str,
+        use_memory: bool = True,
+        memory_config: Optional[dict] = None,
+    ):
+        """
+        Initializes a new Finetune_DialogAgent with specified configuration.
+
+        Arguments:
+            name (str): Name of the agent.
+            sys_prompt (str): System prompt or description of the agent's role.
+            model_config_name (str): The configuration name for
+                                     the underlying model.
+            use_memory (bool, optional): Indicates whether to utilize
+                                         memory features. Defaults to True.
+            memory_config (dict, optional): Configuration for memory
+                                            functionalities if
+                                            `use_memory` is True.
+
+        Note:
+            Refer to `class DialogAgent(AgentBase)` for more information.
+        """
+
+        super().__init__(
+            name,
+            sys_prompt,
+            model_config_name,
+            use_memory,
+            memory_config,
+        )
+
+    def load_model(
+        self,
+        model_id: Optional[str] = None,
+        local_model_path: Optional[str] = None,
+    ) -> None:
+        """
+        Load a new model into the agent.
+
+        Arguments:
+            model_id (str): The Hugging Face model ID or a custom identifier.
+                            Needed if loading model from Hugging Face.
+            local_model_path (str, optional): Path to a locally saved model.
+
+        Raises:
+            Exception: If the model loading process fails or if the
+                       model wrapper does not support dynamic loading.
+        """
+
+        if hasattr(self.model, "load_model"):
+            self.model.load_model(model_id, local_model_path)
+        else:
+            logger.error(
+                "The model wrapper does not support dynamic model loading.",
+            )
+
+    def load_tokenizer(
+        self,
+        model_id: Optional[str] = None,
+        local_tokenizer_path: Optional[str] = None,
+    ) -> None:
+        """
+        Load a new tokenizer for the agent.
+
+        Arguments:
+            model_id (str): The Hugging Face model ID or a custom identifier.
+                            Needed if loading tokenizer from Hugging Face.
+            local_tokenizer_path (str, optional): Path to a locally saved
+                                                  tokenizer.
+
+        Raises:
+            Exception: If the model tokenizer process fails or if the
+                       model wrapper does not support dynamic loading.
+        """
+
+        if hasattr(self.model, "load_tokenizer"):
+            self.model.load_tokenizer(model_id, local_tokenizer_path)
+        else:
+            logger.error("The model wrapper does not support dynamic loading.")
+
+    def fine_tune(
+        self,
+        data_path: Optional[str] = None,
+        fine_tune_config: Optional[Dict[str, Any]] = None,
+    ) -> None:
+        """
+        Fine-tune the agent's underlying model.
+
+        Arguments:
+            data_path (str): The path to the training data.
+
+        Raises:
+            Exception: If fine-tuning fails or if the
+                       model wrapper does not support fine-tuning.
+        """
+
+        if hasattr(self.model, "fine_tune"):
+            self.model.fine_tune(data_path, fine_tune_config)
+            logger.info("Fine-tuning completed successfully.")
+        else:
+            logger.error("The model wrapper does not support fine-tuning.")
diff --git a/examples/load_finetune_huggingface_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
similarity index 77%
rename from examples/load_finetune_huggingface_model/huggingface_model.py
rename to examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 1a95ce7eb..b5c59a4af 100644
--- a/examples/load_finetune_huggingface_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -2,16 +2,9 @@
 """
 This module provides a HuggingFaceWrapper to manage
 and operate Hugging Face Transformers models, enabling loading,
-fine-tuning, and response generation. It includes the
-Finetune_DialogAgent class, which extends DialogAgent to
-enhance fine-tuning capabilities with custom hyperparameters.
+fine-tuning, and response generation.
 Key features include handling model and tokenizer operations,
 adapting to specialized datasets, and robust error management.
-
-Classes:
-- HuggingFaceWrapper: Manages Hugging Face models and tokenizers.
-- Finetune_DialogAgent: Extends DialogAgent for model fine-tuning.
-
 """
 from typing import Sequence, Any, Union, List, Optional, Dict
 import os
@@ -73,7 +66,7 @@ def __init__(
         self.model_id = model_id
         relative_path = os.path.join(
             os.path.dirname(__file__),
-            "../load_finetune_huggingface_model/.env",
+            "../conversation_with_agent_with_finetuned_model/.env",
         )
         dotenv_path = os.path.normpath(relative_path)
         _ = load_dotenv(dotenv_path)  # read local .env file
@@ -144,7 +137,7 @@ def __call__(
             )
             # Decode the generated tokens to a string
             generated_text = self.tokenizer.decode(
-                outputs[0][input_ids.shape[1]:],
+                outputs[0][input_ids.shape[1] :],  # noqa: E203
                 skip_special_tokens=True,
             )
             return ModelResponse(text=generated_text, raw=outputs)
@@ -291,6 +284,7 @@ def load_tokenizer(
                     f"Successfully loaded new tokenizer for model "
                     f"'{model_id}' from Hugging Face",
                 )
+
             else:
                 self.tokenizer = AutoTokenizer.from_pretrained(
                     local_tokenizer_path,
@@ -300,6 +294,7 @@ def load_tokenizer(
                     f"Successfully loaded new tokenizer for model "
                     f"'{model_id}' from '{local_tokenizer_path}'",
                 )
+            self.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
 
         except Exception as e:
             # Handle exceptions during model loading,
@@ -394,6 +389,7 @@ def fine_tune_training(
         import json
 
         dataset = load_dataset(data_path, token=token)
+        dataset = dataset["train"].train_test_split(test_size=0.1)
 
         from peft import LoraConfig
 
@@ -406,24 +402,25 @@ def fine_tune_training(
         }
 
         if fine_tune_config is not None:
-            if fine_tune_config["lora_config"] is not None:
+            if fine_tune_config.get("lora_config") is not None:
                 lora_config_default.update(fine_tune_config["lora_config"])
 
         training_defaults = {
             "per_device_train_batch_size": 1,
-            "gradient_accumulation_steps": 1,
+            # "gradient_accumulation_steps": 1,
             "gradient_checkpointing": False,
-            "max_steps": 10,
+            # "max_steps": 10,
+            "num_train_epochs": 10,
             "output_dir": "./",
             "optim": "paged_adamw_8bit",
             "fp16": True,
             "logging_steps": 1,
-            # "learning_rate": 2e-6,
+            "learning_rate": 1e-5,
             # "num_train_epochs": 10.0,
         }
 
         if fine_tune_config is not None:
-            if fine_tune_config["training_args"] is not None:
+            if fine_tune_config.get("training_args") is not None:
                 training_defaults.update(fine_tune_config["training_args"])
 
         from peft import get_peft_model
@@ -445,25 +442,53 @@ def formatting_prompts_func(
                 output_texts.append(text)
             return output_texts
 
-        response_template = " ### Answer:"
-        collator = DataCollatorForCompletionOnlyLM(
-            response_template,
-            tokenizer=tokenizer,
-        )
+        def formatting_func(example):
+            if example.get("context", "") != "":
+                input_prompt = (
+                    f"Below is an instruction that describes a task, "
+                    f"paired with an input that provides further context. "
+                    f"Write a response that appropriately "
+                    f"completes the request.\n\n"
+                    f"### Instruction:\n"
+                    f"{example['instruction']}\n\n"
+                    f"### Input: \n"
+                    f"{example['context']}\n\n"
+                    f"### Response: \n"
+                    f"{example['response']}"
+                )
+
+            else:
+                input_prompt = (
+                    f"Below is an instruction that describes a task. "
+                    "Write a response that appropriately "
+                    f"completes the request.\n\n"
+                    "### Instruction:\n"
+                    f"{example['instruction']}\n\n"
+                    f"### Response:\n"
+                    f"{example['response']}"
+                )
+
+            return {"text": input_prompt}
+
+        formatted_dataset = dataset.map(formatting_func)
 
         trainer_args = transformers.TrainingArguments(**training_defaults)
 
         trainer = SFTTrainer(
             model,
-            train_dataset=dataset["train"],
-            eval_dataset=dataset["train"],
-            formatting_func=formatting_prompts_func,
-            data_collator=collator,
+            train_dataset=formatted_dataset["train"],
+            eval_dataset=formatted_dataset["test"],
+            # formatting_func=formatting_prompts_func,
+            # data_collator=collator,
             peft_config=lora_config,
             args=trainer_args,
+            dataset_text_field="text",
+            max_seq_length=512,
         )
 
-        print("fine-tuning model")
+        logger.info(
+            "fine-tuning model",
+        )
 
         trainer.train()
 
@@ -498,116 +523,3 @@ def formatting_prompts_func(
         return model
 
 
-class Finetune_DialogAgent(DialogAgent):
-    """
-    A dialog agent capable of fine-tuning its
-    underlying model based on provided data.
-
-    Inherits from DialogAgent and adds functionality for
-    fine-tuning with custom hyperparameters.
-    """
-
-    def __init__(
-        self,
-        name: str,
-        sys_prompt: str,
-        model_config_name: str,
-        use_memory: bool = True,
-        memory_config: Optional[dict] = None,
-    ):
-        """
-        Initializes a new Finetune_DialogAgent with specified configuration.
-
-        Arguments:
-            name (str): Name of the agent.
-            sys_prompt (str): System prompt or description of the agent's role.
-            model_config_name (str): The configuration name for
-                                     the underlying model.
-            use_memory (bool, optional): Indicates whether to utilize
-                                         memory features. Defaults to True.
-            memory_config (dict, optional): Configuration for memory
-                                            functionalities if
-                                            `use_memory` is True.
-
-        Note:
-            Refer to `class DialogAgent(AgentBase)` for more information.
-        """
-
-        super().__init__(
-            name,
-            sys_prompt,
-            model_config_name,
-            use_memory,
-            memory_config,
-        )
-
-    def load_model(
-        self,
-        model_id: Optional[str] = None,
-        local_model_path: Optional[str] = None,
-    ) -> None:
-        """
-        Load a new model into the agent.
-
-        Arguments:
-            model_id (str): The Hugging Face model ID or a custom identifier.
-                            Needed if loading model from Hugging Face.
-            local_model_path (str, optional): Path to a locally saved model.
-
-        Raises:
-            Exception: If the model loading process fails or if the
-                       model wrapper does not support dynamic loading.
-        """
-
-        if hasattr(self.model, "load_model"):
-            self.model.load_model(model_id, local_model_path)
-        else:
-            logger.error(
-                "The model wrapper does not support dynamic model loading.",
-            )
-
-    def load_tokenizer(
-        self,
-        model_id: Optional[str] = None,
-        local_tokenizer_path: Optional[str] = None,
-    ) -> None:
-        """
-        Load a new tokenizer for the agent.
-
-        Arguments:
-            model_id (str): The Hugging Face model ID or a custom identifier.
-                            Needed if loading tokenizer from Hugging Face.
-            local_tokenizer_path (str, optional): Path to a locally saved
-                                                  tokenizer.
-
-        Raises:
-            Exception: If the model tokenizer process fails or if the
-                       model wrapper does not support dynamic loading.
-        """
-
-        if hasattr(self.model, "load_tokenizer"):
-            self.model.load_tokenizer(model_id, local_tokenizer_path)
-        else:
-            logger.error("The model wrapper does not support dynamic loading.")
-
-    def fine_tune(
-        self,
-        data_path: Optional[str] = None,
-        fine_tune_config: Optional[Dict[str, Any]] = None,
-    ) -> None:
-        """
-        Fine-tune the agent's underlying model.
-
-        Arguments:
-            data_path (str): The path to the training data.
-
-        Raises:
-            Exception: If fine-tuning fails or if the
-                       model wrapper does not support fine-tuning.
-        """
-
-        if hasattr(self.model, "fine_tune"):
-            self.model.fine_tune(data_path, fine_tune_config)
-            logger.info("Fine-tuning completed successfully.")
-        else:
-            logger.error("The model wrapper does not support fine-tuning.")
diff --git a/examples/load_finetune_huggingface_model/README.md b/examples/load_finetune_huggingface_model/README.md
deleted file mode 100644
index bdb6172d6..000000000
--- a/examples/load_finetune_huggingface_model/README.md
+++ /dev/null
@@ -1,42 +0,0 @@
-# Multi-Agent Conversation with Custom Model Loading and Fine-Tuning in AgentScope
-
-This example demonstrates how to load and optionally fine-tune a Hugging Face model within a multi-agent conversation setup using AgentScope. The complete code is provided in `load_finetune_huggingface_model.py`.
-
-## Functionality Overview
-
-Compared to basic conversation setup, this example introduces model loading and fine-tuning features:
-
-- Use `dialog_agent.load_model(model_id, local_model_path)` to load a model either from the Hugging Face Model Hub or a local directory.
-- Apply `dialog_agent.fine_tune(data_path)` to fine-tune the model based on your dataset.
-
-The default hyperparameters for (SFT) fine-tuning are specified in `agentscope/src/agentscope/models/huggingface_model.py`. For customized hyperparameters, specify them in `model_configs` if the model needs to be fine-tuned at initialization, or specify through `fine_tune_config` in `Finetune_DialogAgent`'s `fine_tune` method after initialization, as shown in the example script `load_finetune_huggingface_model.py`.
-
-## Agent Initialization
-
-When initializing an agent, the following parameters need specification:
-
-- `model_id` (str): Identifier for the model on Hugging Face.
-- `local_model_path` (str): Local path to the model (defaults to loading from Hugging Face if not provided).
-- `data_path` (str): Path to training data (fine-tuning is skipped if not provided).
-- `device` (str): The device (e.g., 'cuda', 'cpu') for model operation, defaulting to 'cuda' if available.
-- `fine_tune_config` (dict, Optional): A configuration dictionary for fine-tuning the model. It allows specifying hyperparameters and other training options that will be passed to the fine-tuning method. If not provided, default settings will be used. This allows for customization of the fine-tuning process to optimize model performance based on specific requirements.
-- `huggingface_token` (from .env file): Token required for models needing authentication from Hugging Face.
-
-## Tested Models
-
-The example is tested using specific Hugging Face models. While it is designed to be flexible, some models may require additional configuration or modification of the provided scripts.
-
-## Prerequisites
-
-Before running this example, ensure you have installed the following packages:
-
-- `transformers`
-- `peft`
-- `python-dotenv`
-- `datasets`
-- `trl`
-
-Additionally, set your Hugging Face token in the `.env` file:
-
-```bash
-python load_finetune_huggingface_model.py
\ No newline at end of file
diff --git a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py b/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
deleted file mode 100644
index b16f51f26..000000000
--- a/examples/load_finetune_huggingface_model/load_finetune_huggingface_model.py
+++ /dev/null
@@ -1,91 +0,0 @@
-# -*- coding: utf-8 -*-
-"""
-This script sets up a conversational agent using
-AgentScope with a Hugging Face model.
-It includes initializing a Finetune_DialogAgent,
-loading and fine-tuning a pre-trained model,
-and conducting a dialogue via a sequential pipeline.
-The conversation continues until the user exits.
-Features include model and tokenizer loading,
-and fine-tuning on the GAIR/lima dataset with adjustable parameters.
-"""
-from huggingface_model import Finetune_DialogAgent
-
-import agentscope
-from agentscope.agents.user_agent import UserAgent
-from agentscope.pipelines.functional import sequentialpipeline
-
-
-def main() -> None:
-    """A basic conversation demo with a custom model"""
-
-    # Initialize AgentScope with your custom model configuration
-
-    agentscope.init(
-        model_configs=[
-            {
-                "model_type": "huggingface",
-                "config_name": "my_custom_model",
-                # Or another generative model of your choice.
-                # Needed from loading from Hugging Face.
-                "model_id": "google/gemma-2b-it",
-                # "local_model_path":  # Specify your local model path
-                # "local_tokenizer_path":  # Specify your local tokenizer path
-                "max_length": 128,
-                "device": "cuda",
-                # Specify a Hugging Face data path if you
-                # wish to finetune the model from the start
-                "data_path": "GAIR/lima",
-                # fine_tune_config (Optional): Configuration for
-                # fine-tuning the model.
-                # This dictionary can include hyperparameters and other
-                # training options that will be passed to the
-                # fine-tuning method. Defaults to None.
-                "fine_tune_config": {
-                    "lora_config": {"r": 20, "lora_alpha": 40},
-                    "training_args": {"max_steps": 20, "logging_steps": 2},
-                },
-            },
-        ],
-    )
-
-    # Init agents with the custom model
-    dialog_agent = Finetune_DialogAgent(
-        name="Assistant",
-        sys_prompt="You're a helpful assistant.",
-        # Use your custom model config name here
-        model_config_name="my_custom_model",
-    )
-
-    dialog_agent.load_model(
-        model_id="google/gemma-2b-it",
-        local_model_path=None,
-    )  # load model gemma-2b-it from Hugging Face
-    dialog_agent.load_tokenizer(
-        model_id="google/gemma-2b-it",
-        local_tokenizer_path=None,
-    )  # load tokenizer for gemma-2b-it from Hugging Face
-
-    # fine-tune loaded model with lima dataset with default hyperparameters
-    # dialog_agent.fine_tune(data_path=  "GAIR/lima")
-
-    # fine-tune loaded model with lima dataset with customized hyperparameters
-    # (`fine_tune_config` argument is optional. Defaults to None.)
-    dialog_agent.fine_tune(
-        "GAIR/lima",
-        fine_tune_config={
-            "lora_config": {"r": 24, "lora_alpha": 48},
-            "training_args": {"max_steps": 30, "logging_steps": 3},
-        },
-    )
-
-    user_agent = UserAgent()
-
-    # Start the conversation between user and assistant
-    x = None
-    while x is None or x.content != "exit":
-        x = sequentialpipeline([dialog_agent, user_agent], x)
-
-
-if __name__ == "__main__":
-    main()

From 0a079b9b23bf1dc879c3ae35e483ec45036f8524 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 2 May 2024 11:50:19 +0800
Subject: [PATCH 10/55] added updated README

---
 .../README.md                                 | 68 +++++++++++++++++++
 1 file changed, 68 insertions(+)
 create mode 100644 examples/conversation_with_agent_with_finetuned_model/README.md

diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
new file mode 100644
index 000000000..3d68d635f
--- /dev/null
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -0,0 +1,68 @@
+# Multi-Agent Conversation with Custom Model Loading and Fine-Tuning in AgentScope
+
+This example demonstrates how to load and optionally fine-tune a Hugging Face model within a multi-agent conversation setup using AgentScope. The complete code is provided in `agentscope/examples/conversation_with_agent_with_finetuned_model`.
+
+## Functionality Overview
+
+Compared to basic conversation setup, this example introduces model loading and fine-tuning features:
+
+- Use `dialog_agent.load_model(model_id, local_model_path)` to load a model either from the Hugging Face Model Hub or a local directory.
+- Apply `dialog_agent.fine_tune(data_path)` to fine-tune the model based on your dataset.
+
+The default hyperparameters for (SFT) fine-tuning are specified in `agentscope/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py` and `agentscope/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json`. For customized hyperparameters, specify them in `model_configs` if the model needs to be fine-tuned at initialization, or specify through `fine_tune_config` in `Finetune_DialogAgent`'s `fine_tune` method after initialization, as shown in the example script `conversation_with_agent_with_finetuned_model.py`.
+
+## Agent Initialization
+
+When initializing an agent, the following parameters need specification:
+
+- `model_id` (str): Identifier for the model on Hugging Face.
+- `local_model_path` (str): Local path to the model (defaults to loading from Hugging Face if not provided).
+- `data_path` (str): Path to training data (fine-tuning is skipped if not provided).
+- `device` (str): The device (e.g., 'cuda', 'cpu') for model operation, defaulting to 'cuda' if available.
+- `fine_tune_config` (dict, Optional): A configuration dictionary for fine-tuning the model. It allows specifying hyperparameters and other training options that will be passed to the fine-tuning method. If not provided, default settings will be used. This allows for customization of the fine-tuning process to optimize model performance based on specific requirements.
+- `huggingface_token` (from .env file): Token required for models needing authentication from Hugging Face.
+
+## Example Ouputs
+### Before Fine-tuning:
+User:
+
+Explain in simple terms how the attention mechanism of a transformer model works
+
+Assistant:
+
+1. a person who helps another person or group of people. 2. a person who helps another person or group of people.
+
+### After Fine-tuning (with the default configuration in `model_configs`):
+User:
+
+Explain in simple terms how the attention mechanism of a transformer model works
+
+Assistant:
+
+The attention mechanism of a transformer model works by allowing the model to focus on different parts of the input sequence at different times. This is done by using a combination of self-attention and position-wise attention.
+
+The self-attention mechanism allows the model to focus on different parts of the input sequence at different times. This is done by using a combination of self-attention and position-wise attention.
+
+The position-wise attention mechanism allows the model to focus on different parts of the input sequence at different times. This is done by using a combination of self-attention and position-wise attention.
+
+The self
+
+
+## Tested Models
+
+The example is tested using specific Hugging Face model `openlm-research/open_llama_3b_v2` on dataset `databricks/databricks-dolly-15k`. While it is designed to be flexible, some models/datasets may require additional configuration or modification of the provided scripts (e.g., pre-processing of the datasets).
+
+## Prerequisites
+
+Before running this example, ensure you have installed the following packages:
+
+- `transformers`
+- `peft`
+- `python-dotenv`
+- `datasets`
+- `trl`
+
+Additionally, set `HUGGINGFACE_TOKEN` in the `agentscope/examples/conversation_with_agent_with_finetuned_model/.env`.
+
+```bash
+python conversation_with_agent_with_finetuned_model.py
\ No newline at end of file

From a4d1f1beb23d922dc3680347460f07a22f56230b Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Sun, 5 May 2024 20:52:33 +0800
Subject: [PATCH 11/55] updated README for two examples and tested on 3
 model_type.

---
 .../conversation_with_RAG_agents/README.md    | 76 +++++++------------
 examples/conversation_with_mentions/README.md | 75 +++++-------------
 2 files changed, 49 insertions(+), 102 deletions(-)

diff --git a/examples/conversation_with_RAG_agents/README.md b/examples/conversation_with_RAG_agents/README.md
index 5b08379a4..04025f979 100644
--- a/examples/conversation_with_RAG_agents/README.md
+++ b/examples/conversation_with_RAG_agents/README.md
@@ -1,57 +1,37 @@
 # AgentScope Consultants: a Multi-Agent RAG Application
 
-* **What is this example about?**
-With the provided implementation and configuration,
-you will obtain three different agents who can help you answer different questions about AgentScope.
+This example will show
+- How to utilize three different agents to answer various questions about AgentScope.
+- How to set up and run the agents using different configurations.
 
-* **What is this example for?** By this example, we want to show how the agent with retrieval augmented generation (RAG)
-capability can be used to build easily.
+## Background
 
-**Notice:** This example is a Beta version of the AgentScope RAG agent. A formal version will soon be added to `src/agentscope/agents`, but it may be subject to changes.
+This example introduces a multi-agent system using retrieval augmented generation (RAG) capabilities to demonstrate how such systems can be built and used effectively.
+
+## Tested Models
+
+These models are tested in this example. For other models, some modifications may be needed.
+### models
+- dashscope_chat (qwen-max)
+### embeddings
+- dashscope_text_embedding(text-embedding-v2)
 
 ## Prerequisites
-* **Cloning repo:** This example requires cloning the whole AgentScope repo to local.
-* **Packages:** This example is built on the LlamaIndex package. Thus, some packages need to be installed before running the example.
+
+Fill the next cell to meet the following requirements
+- Cloning the AgentScope repository to local.
+- Installation of required packages:
     ```bash
     pip install llama-index tree_sitter tree-sitter-languages
     ```
-* **Model APIs:** This example uses Dashscope APIs. Thus, we also need an API key for DashScope.
-  ```bash
-  export DASH_SCOPE_API='YOUR_API_KEY'
-  ```
-
-**Note:** This example has been tested with `dashscope_chat` and `dashscope_text_embedding` model wrapper, with `qwen-max` and `text-embedding-v2` models.
-However, you are welcome to replace the Dashscope language and embedding model wrappers or models with other models you like to test.
-
-## Start AgentScope Consultants
-* **Terminal:** The most simple way to execute the AgentScope Consultants is running in terminal.
-  ```bash
-  python ./rag_example.py
-  ```
-  Setting `log_retrieval` to `false` in `agent_config.json` can hide the retrieved information and provide only answers of agents.
-
-* **AS studio:** If you want to have more organized, clean UI, you can also run with our `as_studio`.
-  ```bash
-  as_studio ./rag_example.py
-  ```
-
-### Customize AgentScope Consultants to other consultants
-After you run the example, you may notice that this example consists of three RAG agents:
-* `AgentScope Tutorial Assistant`: responsible for answering questions based on AgentScope tutorials (markdown files).
-* `AgentScope Framework Code Assistant`: responsible for answering questions based on AgentScope code base (python files).
-* `Summarize Assistant`: responsible for summarize the questions from the above two agents.
-
-These agents can be configured to answering questions based on other GitHub repo, by simply modifying the `input_dir` fields in the `agent_config.json`.
-
-For more advanced customization, we may need to learn a little bit from the following.
-
-**RAG modules:** In AgentScope, RAG modules are abstract to provide three basic functions: `load_data`, `store_and_index` and `retrieve`. Refer to `src/agentscope/rag` for more details.
-
-**RAG configs:** In the example configuration (the `rag_config` field), all parameters are optional. But if you want to customize them, you may want to learn the following:
-*  `load_data`: contains all parameters for the the `rag.load_data` function.
-Since the `load_data` accepts a dataloader object `loader`, the `loader` in the config need to have `"create_object": true` to let a internal parse create a LlamaIndex data loader object.
-The loader object is an instance of `class` in module `module`, with initialization parameters in `init_args`.
-
-* `store_and_index`: contains all parameters for the the `rag.store_and_index` function.
-For example, you can pass `vector_store` and `retriever` configurations in a similar way as the `loader` mentioned above.
-For the `transformations` parameter, you can pass a list of dicts, each of which corresponds to building a `NodeParser`-kind of preprocessor in Llamaindex.
\ No newline at end of file
+- Setting environment variables for API keys:
+    ```bash
+    export DASH_SCOPE_API='YOUR_API_KEY'
+    ```
+- Running the application via terminal or AS studio:
+    ```bash
+    python ./rag_example.py
+    # or
+    as_studio ./rag_example.py
+    ```
+- [Optional] Optional settings to hide retrieved information by setting `log_retrieval` to `false` in `agent_config.json`.
diff --git a/examples/conversation_with_mentions/README.md b/examples/conversation_with_mentions/README.md
index 6359b3413..2cb12ea52 100644
--- a/examples/conversation_with_mentions/README.md
+++ b/examples/conversation_with_mentions/README.md
@@ -1,73 +1,40 @@
+###
 # Multi-Agent Group Conversation in AgentScope
 
-This example demonstrates a multi-agent group conversation facilitated by AgentScope. The script `main.py` sets up a virtual chat room where a user agent interacts with several NPC (non-player character) agents. The chat utilizes a special **"@"** mention functionality, which allows participants to address specific agents and have a more directed conversation.
+This example demonstrates a multi-agent group conversation facilitated by AgentScope. The script sets up a virtual chat room where a user agent interacts with several NPC (non-player character) agents. Participants can utilize a special "@" mention functionality to address specific agents directly.
 
-## Key Features
+## Background
 
-- **Real-time Group Conversation**: Engage in a chat with multiple agents responding in real time.
-- **@ Mention Functionality**: Use the "@" symbol followed by an agent's name to specifically address that agent within the conversation.
-- **Dynamic Flow**: User-driven conversation with agents responding based on the context and mentions.
-- **Configurable Agent Roles**: Easily modify agent roles and behaviors by editing the `sys_prompt` in the configuration files.
-- **User Timeout**: If the user does not respond within a specified time, the conversation continues with the next agent.
+The conversation takes place in a simulated chat room environment with predefined roles for each participant. Topics are open-ended and evolve based on the user's input and agents' responses.
 
-## How to Use
-
-To start the group conversation, follow these steps:
-
-1. Make sure to set your `api_key` in the `configs/model_configs.json` file.
-2. Run the script using the following command:
+## Tested Models
 
-```bash
-python main.py
+These models are tested in this example. For other models, some modifications may be needed.
+- gemini_chat (models/gemini-pro, models/gemini-1.0-pro)
+- dashscope_chat (qwen-max, qwen-turbo)
+- ollama_chat (ollama_llama3_8b)
 
-# or launch agentscope studio
-as_studio main.py
-```
+## Prerequisites
 
-1. To address a specific agent in the chat, type "@" followed by the agent's name in your message.
-2. To exit the chat, simply type "exit" when it's your turn to speak.
+Fill the next cell to meet the following requirements:
+- Set your `api_key` in the `configs/model_configs.json` file
+- Optional: Launch agentscope studio with `as_studio main.py`
 
-## Background and Conversation Flow
-
-The conversation takes place in a simulated chat room environment with roles defined for each participant. The user acts as a regular chat member with the ability to speak freely and address any agent. NPC agents are pre-configured with specific roles that determine their responses and behavior in the chat. The topic of the conversation is open-ended and can evolve organically based on the user's input and agents' programmed personas.
-
-### Example Interaction
+## How to Use
 
-```
-User input: Hi, everyone! I'm excited to join this chat.
-AgentA: Welcome! We're glad to have you here.
-User input: @AgentB, what do you think about the new technology trends?
-AgentB: It's an exciting time for tech! There are so many innovations on the horizon.
-...
-```
+1. Run the script using the command: `python main.py`
+2. Address specific agents by typing "@" followed by the agent's name.
+3. Type "exit" to leave the chat.
 
 ## Customization Options
 
-The group conversation script provides several options for customization, allowing you to tailor the chat experience to your preferences.
-
-You can customize the conversation by editing the agent configurations and model parameters. The `agent_configs.json` file allows you to set specific behaviors for each NPC agent, while `model_configs.json` contains the parameters for the conversation model.
+You can adjust the behavior and parameters of the NPC agents and conversation model by editing the `agent_configs.json` and `model_configs.json` files, respectively.
 
 ### Changing User Input Time Limit
 
-The `USER_TIME_TO_SPEAK` variable sets the time limit (in seconds) for the user to input their message during each round. By default, this is set to 10 seconds. You can adjust this time limit by modifying the value of `USER_TIME_TO_SPEAK` in the `main.py` script.
-
-For example, to change the time limit to 20 seconds, update the line in `main.py` as follows:
-
-```
-USER_TIME_TO_SPEAK = 20  # User has 20 seconds to type their message
-```
+Adjust the `USER_TIME_TO_SPEAK` variable in the `main.py` script to change the time limit for user input.
 
 ### Setting a Default Topic for the Chat Room
 
-The `DEFAULT_TOPIC` variable defines the initial message or topic of the chat room. It sets the stage for the conversation and is announced at the beginning of the chat session. You can change this message to prompt a specific discussion topic or to provide instructions to the agents.
-
-To customize this message, modify the `DEFAULT_TOPIC` variable in the `main.py` script. For instance, if you want to set the default topic to discuss "The Future of Artificial Intelligence," you would change the code as follows:
-
-```python
-DEFAULT_TOPIC = """
-This is a chat room about the Future of Artificial Intelligence and you can
-speak freely and briefly.
-"""
-```
-
-With these customizations, the chat room can be tailored to fit specific themes or time constraints, enhancing the user's control over the chat experience.
+Modify the `DEFAULT_TOPIC` variable in the `main.py` script to set the initial topic of the chat room.
+###
\ No newline at end of file

From 6b5410edd50e61aa7c08f0dd5d7ae9069c90711f Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Mon, 6 May 2024 11:09:17 +0800
Subject: [PATCH 12/55] undo update to conversation_with_mentions README
 (created a dedicated branch for it)

---
 examples/conversation_with_mentions/README.md | 75 +++++++++++++------
 1 file changed, 54 insertions(+), 21 deletions(-)

diff --git a/examples/conversation_with_mentions/README.md b/examples/conversation_with_mentions/README.md
index 2cb12ea52..6359b3413 100644
--- a/examples/conversation_with_mentions/README.md
+++ b/examples/conversation_with_mentions/README.md
@@ -1,40 +1,73 @@
-###
 # Multi-Agent Group Conversation in AgentScope
 
-This example demonstrates a multi-agent group conversation facilitated by AgentScope. The script sets up a virtual chat room where a user agent interacts with several NPC (non-player character) agents. Participants can utilize a special "@" mention functionality to address specific agents directly.
+This example demonstrates a multi-agent group conversation facilitated by AgentScope. The script `main.py` sets up a virtual chat room where a user agent interacts with several NPC (non-player character) agents. The chat utilizes a special **"@"** mention functionality, which allows participants to address specific agents and have a more directed conversation.
 
-## Background
+## Key Features
 
-The conversation takes place in a simulated chat room environment with predefined roles for each participant. Topics are open-ended and evolve based on the user's input and agents' responses.
+- **Real-time Group Conversation**: Engage in a chat with multiple agents responding in real time.
+- **@ Mention Functionality**: Use the "@" symbol followed by an agent's name to specifically address that agent within the conversation.
+- **Dynamic Flow**: User-driven conversation with agents responding based on the context and mentions.
+- **Configurable Agent Roles**: Easily modify agent roles and behaviors by editing the `sys_prompt` in the configuration files.
+- **User Timeout**: If the user does not respond within a specified time, the conversation continues with the next agent.
 
-## Tested Models
+## How to Use
 
-These models are tested in this example. For other models, some modifications may be needed.
-- gemini_chat (models/gemini-pro, models/gemini-1.0-pro)
-- dashscope_chat (qwen-max, qwen-turbo)
-- ollama_chat (ollama_llama3_8b)
+To start the group conversation, follow these steps:
 
-## Prerequisites
+1. Make sure to set your `api_key` in the `configs/model_configs.json` file.
+2. Run the script using the following command:
 
-Fill the next cell to meet the following requirements:
-- Set your `api_key` in the `configs/model_configs.json` file
-- Optional: Launch agentscope studio with `as_studio main.py`
+```bash
+python main.py
 
-## How to Use
+# or launch agentscope studio
+as_studio main.py
+```
+
+1. To address a specific agent in the chat, type "@" followed by the agent's name in your message.
+2. To exit the chat, simply type "exit" when it's your turn to speak.
+
+## Background and Conversation Flow
+
+The conversation takes place in a simulated chat room environment with roles defined for each participant. The user acts as a regular chat member with the ability to speak freely and address any agent. NPC agents are pre-configured with specific roles that determine their responses and behavior in the chat. The topic of the conversation is open-ended and can evolve organically based on the user's input and agents' programmed personas.
+
+### Example Interaction
 
-1. Run the script using the command: `python main.py`
-2. Address specific agents by typing "@" followed by the agent's name.
-3. Type "exit" to leave the chat.
+```
+User input: Hi, everyone! I'm excited to join this chat.
+AgentA: Welcome! We're glad to have you here.
+User input: @AgentB, what do you think about the new technology trends?
+AgentB: It's an exciting time for tech! There are so many innovations on the horizon.
+...
+```
 
 ## Customization Options
 
-You can adjust the behavior and parameters of the NPC agents and conversation model by editing the `agent_configs.json` and `model_configs.json` files, respectively.
+The group conversation script provides several options for customization, allowing you to tailor the chat experience to your preferences.
+
+You can customize the conversation by editing the agent configurations and model parameters. The `agent_configs.json` file allows you to set specific behaviors for each NPC agent, while `model_configs.json` contains the parameters for the conversation model.
 
 ### Changing User Input Time Limit
 
-Adjust the `USER_TIME_TO_SPEAK` variable in the `main.py` script to change the time limit for user input.
+The `USER_TIME_TO_SPEAK` variable sets the time limit (in seconds) for the user to input their message during each round. By default, this is set to 10 seconds. You can adjust this time limit by modifying the value of `USER_TIME_TO_SPEAK` in the `main.py` script.
+
+For example, to change the time limit to 20 seconds, update the line in `main.py` as follows:
+
+```
+USER_TIME_TO_SPEAK = 20  # User has 20 seconds to type their message
+```
 
 ### Setting a Default Topic for the Chat Room
 
-Modify the `DEFAULT_TOPIC` variable in the `main.py` script to set the initial topic of the chat room.
-###
\ No newline at end of file
+The `DEFAULT_TOPIC` variable defines the initial message or topic of the chat room. It sets the stage for the conversation and is announced at the beginning of the chat session. You can change this message to prompt a specific discussion topic or to provide instructions to the agents.
+
+To customize this message, modify the `DEFAULT_TOPIC` variable in the `main.py` script. For instance, if you want to set the default topic to discuss "The Future of Artificial Intelligence," you would change the code as follows:
+
+```python
+DEFAULT_TOPIC = """
+This is a chat room about the Future of Artificial Intelligence and you can
+speak freely and briefly.
+"""
+```
+
+With these customizations, the chat room can be tailored to fit specific themes or time constraints, enhancing the user's control over the chat experience.

From 6d100512153a8bc3b8be7cc68b91a002af08f2a8 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Mon, 6 May 2024 11:12:25 +0800
Subject: [PATCH 13/55] reverted changes made to
 conversation_with_RAG_agents\README.md

---
 .../conversation_with_RAG_agents/README.md    | 76 ++++++++++++-------
 1 file changed, 48 insertions(+), 28 deletions(-)

diff --git a/examples/conversation_with_RAG_agents/README.md b/examples/conversation_with_RAG_agents/README.md
index 04025f979..5b08379a4 100644
--- a/examples/conversation_with_RAG_agents/README.md
+++ b/examples/conversation_with_RAG_agents/README.md
@@ -1,37 +1,57 @@
 # AgentScope Consultants: a Multi-Agent RAG Application
 
-This example will show
-- How to utilize three different agents to answer various questions about AgentScope.
-- How to set up and run the agents using different configurations.
+* **What is this example about?**
+With the provided implementation and configuration,
+you will obtain three different agents who can help you answer different questions about AgentScope.
 
-## Background
+* **What is this example for?** By this example, we want to show how the agent with retrieval augmented generation (RAG)
+capability can be used to build easily.
 
-This example introduces a multi-agent system using retrieval augmented generation (RAG) capabilities to demonstrate how such systems can be built and used effectively.
-
-## Tested Models
-
-These models are tested in this example. For other models, some modifications may be needed.
-### models
-- dashscope_chat (qwen-max)
-### embeddings
-- dashscope_text_embedding(text-embedding-v2)
+**Notice:** This example is a Beta version of the AgentScope RAG agent. A formal version will soon be added to `src/agentscope/agents`, but it may be subject to changes.
 
 ## Prerequisites
-
-Fill the next cell to meet the following requirements
-- Cloning the AgentScope repository to local.
-- Installation of required packages:
+* **Cloning repo:** This example requires cloning the whole AgentScope repo to local.
+* **Packages:** This example is built on the LlamaIndex package. Thus, some packages need to be installed before running the example.
     ```bash
     pip install llama-index tree_sitter tree-sitter-languages
     ```
-- Setting environment variables for API keys:
-    ```bash
-    export DASH_SCOPE_API='YOUR_API_KEY'
-    ```
-- Running the application via terminal or AS studio:
-    ```bash
-    python ./rag_example.py
-    # or
-    as_studio ./rag_example.py
-    ```
-- [Optional] Optional settings to hide retrieved information by setting `log_retrieval` to `false` in `agent_config.json`.
+* **Model APIs:** This example uses Dashscope APIs. Thus, we also need an API key for DashScope.
+  ```bash
+  export DASH_SCOPE_API='YOUR_API_KEY'
+  ```
+
+**Note:** This example has been tested with `dashscope_chat` and `dashscope_text_embedding` model wrapper, with `qwen-max` and `text-embedding-v2` models.
+However, you are welcome to replace the Dashscope language and embedding model wrappers or models with other models you like to test.
+
+## Start AgentScope Consultants
+* **Terminal:** The most simple way to execute the AgentScope Consultants is running in terminal.
+  ```bash
+  python ./rag_example.py
+  ```
+  Setting `log_retrieval` to `false` in `agent_config.json` can hide the retrieved information and provide only answers of agents.
+
+* **AS studio:** If you want to have more organized, clean UI, you can also run with our `as_studio`.
+  ```bash
+  as_studio ./rag_example.py
+  ```
+
+### Customize AgentScope Consultants to other consultants
+After you run the example, you may notice that this example consists of three RAG agents:
+* `AgentScope Tutorial Assistant`: responsible for answering questions based on AgentScope tutorials (markdown files).
+* `AgentScope Framework Code Assistant`: responsible for answering questions based on AgentScope code base (python files).
+* `Summarize Assistant`: responsible for summarize the questions from the above two agents.
+
+These agents can be configured to answering questions based on other GitHub repo, by simply modifying the `input_dir` fields in the `agent_config.json`.
+
+For more advanced customization, we may need to learn a little bit from the following.
+
+**RAG modules:** In AgentScope, RAG modules are abstract to provide three basic functions: `load_data`, `store_and_index` and `retrieve`. Refer to `src/agentscope/rag` for more details.
+
+**RAG configs:** In the example configuration (the `rag_config` field), all parameters are optional. But if you want to customize them, you may want to learn the following:
+*  `load_data`: contains all parameters for the the `rag.load_data` function.
+Since the `load_data` accepts a dataloader object `loader`, the `loader` in the config need to have `"create_object": true` to let a internal parse create a LlamaIndex data loader object.
+The loader object is an instance of `class` in module `module`, with initialization parameters in `init_args`.
+
+* `store_and_index`: contains all parameters for the the `rag.store_and_index` function.
+For example, you can pass `vector_store` and `retriever` configurations in a similar way as the `loader` mentioned above.
+For the `transformations` parameter, you can pass a list of dicts, each of which corresponds to building a `NodeParser`-kind of preprocessor in Llamaindex.
\ No newline at end of file

From db27edd79285b5046c31ba7b00ab1f929fff8239 Mon Sep 17 00:00:00 2001
From: Zhang Ze Yu <zyzhang1130@gmail.com>
Date: Mon, 6 May 2024 04:42:47 +0000
Subject: [PATCH 14/55] resolved pre-commit related issues

---
 .../configs/model_configs.json                | 34 ++++-----
 ...rsation_with_agent_with_finetuned_model.py | 76 ++++++++++---------
 .../finetune_dialogagent.py                   | 14 +---
 .../huggingface_model.py                      | 20 +----
 4 files changed, 62 insertions(+), 82 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json b/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
index 3b59c8318..ed491e572 100644
--- a/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
+++ b/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
@@ -1,18 +1,18 @@
-[
-    {
-        "model_type": "huggingface",
-        "config_name": "my_custom_model",
-
-        "model_id": "openlm-research/open_llama_3b_v2",
-
-        "max_length": 128,
-        "device": "cuda",
-
-        "data_path": "databricks/databricks-dolly-15k",
-
-        "fine_tune_config": {
-            "lora_config": {"r": 20, "lora_alpha": 40},
-            "training_args": {"max_steps": 1000, "logging_steps": 1}
-        }
-    }
+[
+    {
+        "model_type": "huggingface",
+        "config_name": "my_custom_model",
+
+        "model_id": "openlm-research/open_llama_3b_v2",
+
+        "max_length": 128,
+        "device": "cuda",
+
+        "data_path": "databricks/databricks-dolly-15k",
+
+        "fine_tune_config": {
+            "lora_config": {"r": 20, "lora_alpha": 40},
+            "training_args": {"max_steps": 1000, "logging_steps": 1}
+        }
+    }
 ]
\ No newline at end of file
diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 989e8525f..4e318b796 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -10,7 +10,7 @@
 and fine-tuning on the databricks-dolly-15k dataset with adjustable parameters.
 """
 from finetune_dialogagent import Finetune_DialogAgent
-
+from huggingface_model import HuggingFaceWrapper
 import agentscope
 from agentscope.agents.user_agent import UserAgent
 from agentscope.pipelines.functional import sequentialpipeline
@@ -21,41 +21,41 @@ def main() -> None:
 
     # Initialize AgentScope with your custom model configuration
 
-    # agentscope.init(
-    #     model_configs=[
-    #         {
-    #             "model_type": "huggingface",
-    #             "config_name": "my_custom_model",
-    #             # Or another generative model of your choice.
-    #             # Needed from loading from Hugging Face.
-    #             "model_id": "openlm-research/open_llama_3b_v2",
-    #             # "local_model_path":  # Specify your local model path
-    #             # "local_tokenizer_path":  # Specify your local tokenizer path
-    #             "max_length": 128,
-    #             "device": "cuda",
-    #             # Specify a Hugging Face data path if you
-    #             # wish to finetune the model from the start
-    #             "data_path": "databricks/databricks-dolly-15k",
-    #             # fine_tune_config (Optional): Configuration for
-    #             # fine-tuning the model.
-    #             # This dictionary can include hyperparameters and other
-    #             # training options that will be passed to the
-    #             # fine-tuning method. Defaults to None.
-    #             # `lora_config` and `training_args` follow 
-    #             # the standard lora and sfttrainer fields.
-    #             "fine_tune_config": {
-    #                 "lora_config": {"r": 20, "lora_alpha": 40},
-    #                 "training_args": {"max_steps": 1000, "logging_steps": 1},
-    #             },
-    #         },
-    #     ],
-    # )
-
-    # alternatively can load `model_configs` from json file
     agentscope.init(
-    model_configs="./configs/model_configs.json",
+        model_configs=[
+            {
+                "model_type": "huggingface",
+                "config_name": "my_custom_model",
+                # Or another generative model of your choice.
+                # Needed from loading from Hugging Face.
+                "model_id": "openlm-research/open_llama_3b_v2",
+                # "local_model_path":  # Specify your local model path
+                # "local_tokenizer_path":  # Specify your local tokenizer path
+                "max_length": 128,
+                "device": "cuda",
+                # Specify a Hugging Face data path if you
+                # wish to finetune the model from the start
+                "data_path": "databricks/databricks-dolly-15k",
+                # fine_tune_config (Optional): Configuration for
+                # fine-tuning the model.
+                # This dictionary can include hyperparameters and other
+                # training options that will be passed to the
+                # fine-tuning method. Defaults to None.
+                # `lora_config` and `training_args` follow
+                # the standard lora and sfttrainer fields.
+                "fine_tune_config": {
+                    "lora_config": {"r": 20, "lora_alpha": 40},
+                    "training_args": {"max_steps": 1000, "logging_steps": 1},
+                },
+            },
+        ],
     )
 
+    # # alternatively can load `model_configs` from json file
+    # agentscope.init(
+    #     model_configs="./configs/model_configs.json",
+    # )
+
     # Init agents with the custom model
     dialog_agent = Finetune_DialogAgent(
         name="Assistant",
@@ -73,16 +73,18 @@ def main() -> None:
         local_tokenizer_path=None,
     )  # load tokenizer for gemma-2b-it from Hugging Face
 
-    # fine-tune loaded model with databricks-dolly-15k dataset with default hyperparameters
-    dialog_agent.fine_tune(data_path="databricks/databricks-dolly-15k")
+    # fine-tune loaded model with databricks-dolly-15k dataset
+    # with default hyperparameters
+    # dialog_agent.fine_tune(data_path="databricks/databricks-dolly-15k")
 
-    # fine-tune loaded model with databricks-dolly-15k dataset with customized hyperparameters
+    # fine-tune loaded model with databricks-dolly-15k dataset
+    # with customized hyperparameters
     # (`fine_tune_config` argument is optional. Defaults to None.)
     dialog_agent.fine_tune(
         "databricks/databricks-dolly-15k",
         fine_tune_config={
             "lora_config": {"r": 24, "lora_alpha": 48},
-            "training_args": {"max_steps": 30, "logging_steps": 3},
+            "training_args": {"max_steps": 300, "logging_steps": 3},
         },
     )
 
diff --git a/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py b/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
index 1059cc99e..10c5c583f 100644
--- a/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
+++ b/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
@@ -1,21 +1,15 @@
 # -*- coding: utf-8 -*-
 """
-This module provides the Finetune_DialogAgent class, 
-which extends DialogAgent to enhance fine-tuning 
+This module provides the Finetune_DialogAgent class,
+which extends DialogAgent to enhance fine-tuning
 capabilities with custom hyperparameters.
 """
-from typing import Sequence, Any, Union, List, Optional, Dict
-import os
+from typing import Any, Optional, Dict
 
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer
 from loguru import logger
-from dotenv import load_dotenv
 
 from agentscope.agents import DialogAgent
-from agentscope.models import ModelWrapperBase, ModelResponse
-from agentscope.message import MessageBase
-from agentscope.utils.tools import _convert_to_str
+
 
 class Finetune_DialogAgent(DialogAgent):
     """
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index b5c59a4af..a7f86ef4b 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -14,7 +14,6 @@
 from loguru import logger
 from dotenv import load_dotenv
 
-from agentscope.agents import DialogAgent
 from agentscope.models import ModelWrapperBase, ModelResponse
 from agentscope.message import MessageBase
 from agentscope.utils.tools import _convert_to_str
@@ -428,21 +427,10 @@ def fine_tune_training(
         lora_config = LoraConfig(**lora_config_default)
         model = get_peft_model(model, lora_config)
 
-        from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
+        from trl import SFTTrainer
         import transformers
 
-        def formatting_prompts_func(
-            example: Dict[str, List[List[str]]],
-        ) -> List[str]:
-            output_texts = []
-            for i in range(len(example["conversations"])):
-                question = f"### Question: {example['conversations'][i][0]}"
-                answer = f"### Answer: {example['conversations'][i][1]}"
-                text = f"{question}\n {answer}"
-                output_texts.append(text)
-            return output_texts
-
-        def formatting_func(example):
+        def formatting_func(example: Dict[str, Any]) -> Dict[str, str]:
             if example.get("context", "") != "":
                 input_prompt = (
                     f"Below is an instruction that describes a task, "
@@ -456,7 +444,6 @@ def formatting_func(example):
                     f"### Response: \n"
                     f"{example['response']}"
                 )
-
             else:
                 input_prompt = (
                     f"Below is an instruction that describes a task. "
@@ -478,7 +465,6 @@ def formatting_func(example):
             model,
             train_dataset=formatted_dataset["train"],
             eval_dataset=formatted_dataset["test"],
-            # formatting_func=formatting_prompts_func,
             # data_collator=collator,
             peft_config=lora_config,
             args=trainer_args,
@@ -521,5 +507,3 @@ def formatting_func(example):
         tokenizer.save_pretrained(tokenizer_path)
 
         return model
-
-

From b37122655d19b6dd1b25f607bfaddb0c64456a7a Mon Sep 17 00:00:00 2001
From: Zhang Ze Yu <zyzhang1130@gmail.com>
Date: Mon, 6 May 2024 05:03:07 +0000
Subject: [PATCH 15/55] resolved pre-commit related issues

---
 .../conversation_with_agent_with_finetuned_model.py           | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 4e318b796..aad691502 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -10,7 +10,9 @@
 and fine-tuning on the databricks-dolly-15k dataset with adjustable parameters.
 """
 from finetune_dialogagent import Finetune_DialogAgent
-from huggingface_model import HuggingFaceWrapper
+from huggingface_model import (
+    HuggingFaceWrapper,
+)  # pylint: disable=unused-import
 import agentscope
 from agentscope.agents.user_agent import UserAgent
 from agentscope.pipelines.functional import sequentialpipeline

From 7f3a0129097fbe553ad86f55539d5f18e41cb054 Mon Sep 17 00:00:00 2001
From: Zhang Ze Yu <zyzhang1130@gmail.com>
Date: Mon, 6 May 2024 05:20:01 +0000
Subject: [PATCH 16/55] resolved pre-commit related issues

---
 .../conversation_with_agent_with_finetuned_model.py         | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index aad691502..5c9836bd0 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -10,9 +10,9 @@
 and fine-tuning on the databricks-dolly-15k dataset with adjustable parameters.
 """
 from finetune_dialogagent import Finetune_DialogAgent
-from huggingface_model import (
-    HuggingFaceWrapper,
-)  # pylint: disable=unused-import
+
+# pylint: disable=unused-import
+from huggingface_model import HuggingFaceWrapper
 import agentscope
 from agentscope.agents.user_agent import UserAgent
 from agentscope.pipelines.functional import sequentialpipeline

From 15bf79ad42fb9e81016f0d820477b89e0da0940d Mon Sep 17 00:00:00 2001
From: Zhang Ze Yu <zyzhang1130@gmail.com>
Date: Wed, 8 May 2024 17:04:20 +0000
Subject: [PATCH 17/55] resolve issues mentioned

---
 .../en/source/tutorial/105-logging.md         |   2 +-
 .../zh_CN/source/tutorial/105-logging.md      |   2 +-
 .../README.md                                 |  42 +--
 .../configs/model_configs.json                |  12 +-
 ...rsation_with_agent_with_finetuned_model.py |  48 ++--
 .../finetune_dialogagent.py                   |  30 ++-
 .../huggingface_model.py                      | 248 +++++++++++++-----
 examples/game_gomoku/code/board_agent.py      |   6 +-
 examples/game_gomoku/main.ipynb               |   6 +-
 src/agentscope/web/README.md                  |   2 +-
 10 files changed, 271 insertions(+), 127 deletions(-)

diff --git a/docs/sphinx_doc/en/source/tutorial/105-logging.md b/docs/sphinx_doc/en/source/tutorial/105-logging.md
index 98f872a8b..7575b11cd 100644
--- a/docs/sphinx_doc/en/source/tutorial/105-logging.md
+++ b/docs/sphinx_doc/en/source/tutorial/105-logging.md
@@ -75,7 +75,7 @@ You can run the WebUI in the following python code:
 import agentscope
 
 agentscope.web.init(
-    path_save="YOUR_SAVE_PATH"
+    path_save="YOUR_output_dir"
 )
 ```
 
diff --git a/docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md b/docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md
index d517cd46b..073d30515 100644
--- a/docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md
+++ b/docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md
@@ -76,7 +76,7 @@ logger.error("The agent encountered an unexpected error while processing a reque
 import agentscope
 
 agentscope.web.init(
-    path_save="YOUR_SAVE_PATH"
+    path_save="YOUR_output_dir"
 )
 ```
 
diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
index 3d68d635f..366f2842c 100644
--- a/examples/conversation_with_agent_with_finetuned_model/README.md
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -6,8 +6,8 @@ This example demonstrates how to load and optionally fine-tune a Hugging Face mo
 
 Compared to basic conversation setup, this example introduces model loading and fine-tuning features:
 
-- Use `dialog_agent.load_model(model_id, local_model_path)` to load a model either from the Hugging Face Model Hub or a local directory.
-- Apply `dialog_agent.fine_tune(data_path)` to fine-tune the model based on your dataset.
+- Initialize an agent or use `dialog_agent.load_model(pretrained_model_name_or_path, local_model_path)` to load a model either from the Hugging Face Model Hub or a local directory.
+- Initalize an agent or apply `dialog_agent.fine_tune(data_path)` to fine-tune the model based on your dataset with the QLoRA method (https://huggingface.co/blog/4bit-transformers-bitsandbytes).
 
 The default hyperparameters for (SFT) fine-tuning are specified in `agentscope/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py` and `agentscope/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json`. For customized hyperparameters, specify them in `model_configs` if the model needs to be fine-tuned at initialization, or specify through `fine_tune_config` in `Finetune_DialogAgent`'s `fine_tune` method after initialization, as shown in the example script `conversation_with_agent_with_finetuned_model.py`.
 
@@ -15,7 +15,7 @@ The default hyperparameters for (SFT) fine-tuning are specified in `agentscope/e
 
 When initializing an agent, the following parameters need specification:
 
-- `model_id` (str): Identifier for the model on Hugging Face.
+- `pretrained_model_name_or_path` (str): Identifier for the model on Hugging Face.
 - `local_model_path` (str): Local path to the model (defaults to loading from Hugging Face if not provided).
 - `data_path` (str): Path to training data (fine-tuning is skipped if not provided).
 - `device` (str): The device (e.g., 'cuda', 'cpu') for model operation, defaulting to 'cuda' if available.
@@ -24,33 +24,39 @@ When initializing an agent, the following parameters need specification:
 
 ## Example Ouputs
 ### Before Fine-tuning:
-User:
+```
+System:
 
-Explain in simple terms how the attention mechanism of a transformer model works
+Explain in simple terms how the attention mechanism of a transformer model works.
 
 Assistant:
 
-1. a person who helps another person or group of people. 2. a person who helps another person or group of people.
-
+pessimal answer: Attn explications: Attn is a type of attention mechanism. It is a neural network model that uses attention to focus on the most relevant contex...
+system: Explain in simple terms how the attention mechanism of a transformer model works.
+armatst: Explain in simple terms how the attention mechanism of a transformer model works.
+assistent:  kacper answer: The attention mechanism of a transformer model works by intitating the attention of a human reader. It glances at the contex...
+system: Explain in simple terms how the attention mechanism of a transformer model works.
+assistent: Explain in simple terms how the
+```
 ### After Fine-tuning (with the default configuration in `model_configs`):
-User:
+```
+System:
 
-Explain in simple terms how the attention mechanism of a transformer model works
+Explain in simple terms how the attention mechanism of a transformer model works.
 
 Assistant:
 
-The attention mechanism of a transformer model works by allowing the model to focus on different parts of the input sequence at different times. This is done by using a combination of self-attention and position-wise attention.
-
-The self-attention mechanism allows the model to focus on different parts of the input sequence at different times. This is done by using a combination of self-attention and position-wise attention.
-
-The position-wise attention mechanism allows the model to focus on different parts of the input sequence at different times. This is done by using a combination of self-attention and position-wise attention.
-
-The self
-
+Sure, the attention mechanism of a transformer model is an important part of the model's ability to generate coherent text. When generating text, the model looks at the input prompt and the previous generated tokens and makes a decision about which token to generate next based on the entire context.
+Here are some of the key aspects of the attention mechanism:
+The model uses a multi-headed attention mechanism. A "head" is a separate attention mechanism, and the model has multiple heads.
+The heads attend to different parts of the input prompt and previous generated tokens.
+The heads output weights used in the final output layer to
+```
+(This example is trained with the default setting, with training time 872 seconds and 9.914 GB gpu memory cost. Reduce training batch size can reduce the memory required. Note that the model is loaded in 4 bits (i.e., QLoRA)).
 
 ## Tested Models
 
-The example is tested using specific Hugging Face model `openlm-research/open_llama_3b_v2` on dataset `databricks/databricks-dolly-15k`. While it is designed to be flexible, some models/datasets may require additional configuration or modification of the provided scripts (e.g., pre-processing of the datasets).
+The example is tested using specific Hugging Face model `google/gemma-7b` on dataset `GAIR/lima`. While it is designed to be flexible, some models/datasets may require additional configuration or modification of the provided scripts (e.g., pre-processing of the datasets in `agentscope/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py`).
 
 ## Prerequisites
 
diff --git a/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json b/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
index ed491e572..ff105c00c 100644
--- a/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
+++ b/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json
@@ -3,16 +3,20 @@
         "model_type": "huggingface",
         "config_name": "my_custom_model",
 
-        "model_id": "openlm-research/open_llama_3b_v2",
+        "pretrained_model_name_or_path": "google/gemma-7b",
 
         "max_length": 128,
         "device": "cuda",
 
-        "data_path": "databricks/databricks-dolly-15k",
+        "data_path": "GAIR/lima",
 
         "fine_tune_config": {
-            "lora_config": {"r": 20, "lora_alpha": 40},
-            "training_args": {"max_steps": 1000, "logging_steps": 1}
+            "lora_config": {"r": 16, "lora_alpha": 32},
+            "training_args": {"max_steps": 200, "logging_steps": 1},
+            "bnb_config" : {"load_in_4bit": "True",
+                                    "bnb_4bit_use_double_quant": "True",
+                                    "bnb_4bit_quant_type": "nf4",
+                                    "bnb_4bit_compute_dtype": "torch.bfloat16"}
         }
     }
 ]
\ No newline at end of file
diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 5c9836bd0..9ec04c545 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -7,12 +7,11 @@
 and conducting a dialogue via a sequential pipeline.
 The conversation continues until the user exits.
 Features include model and tokenizer loading,
-and fine-tuning on the databricks-dolly-15k dataset with adjustable parameters.
+and fine-tuning on the lima dataset with adjustable parameters.
 """
-from finetune_dialogagent import Finetune_DialogAgent
-
 # pylint: disable=unused-import
 from huggingface_model import HuggingFaceWrapper
+from finetune_dialogagent import Finetune_DialogAgent
 import agentscope
 from agentscope.agents.user_agent import UserAgent
 from agentscope.pipelines.functional import sequentialpipeline
@@ -30,14 +29,16 @@ def main() -> None:
                 "config_name": "my_custom_model",
                 # Or another generative model of your choice.
                 # Needed from loading from Hugging Face.
-                "model_id": "openlm-research/open_llama_3b_v2",
+                "pretrained_model_name_or_path": "google/gemma-7b",
                 # "local_model_path":  # Specify your local model path
                 # "local_tokenizer_path":  # Specify your local tokenizer path
                 "max_length": 128,
+                # Device for inference. Fine-tuning occurs on gpus.
                 "device": "cuda",
                 # Specify a Hugging Face data path if you
                 # wish to finetune the model from the start
-                "data_path": "databricks/databricks-dolly-15k",
+                "data_path": "GAIR/lima",
+                # "output_dir":
                 # fine_tune_config (Optional): Configuration for
                 # fine-tuning the model.
                 # This dictionary can include hyperparameters and other
@@ -46,8 +47,14 @@ def main() -> None:
                 # `lora_config` and `training_args` follow
                 # the standard lora and sfttrainer fields.
                 "fine_tune_config": {
-                    "lora_config": {"r": 20, "lora_alpha": 40},
-                    "training_args": {"max_steps": 1000, "logging_steps": 1},
+                    "lora_config": {"r": 16, "lora_alpha": 32},
+                    "training_args": {"max_steps": 200, "logging_steps": 1},
+                    "bnb_config": {
+                        "load_in_4bit": True,
+                        "bnb_4bit_use_double_quant": True,
+                        "bnb_4bit_quant_type": "nf4",
+                        "bnb_4bit_compute_dtype": "torch.bfloat16",
+                    },
                 },
             },
         ],
@@ -61,34 +68,39 @@ def main() -> None:
     # Init agents with the custom model
     dialog_agent = Finetune_DialogAgent(
         name="Assistant",
-        sys_prompt="You're a helpful assistant.",
+        sys_prompt=(
+            "Explain in simple terms how the attention mechanism of "
+            "a transformer model works."
+        ),
         # Use your custom model config name here
         model_config_name="my_custom_model",
     )
 
+    # (Optional) can load another model after
+    # the agent has been instantiated if needed
     dialog_agent.load_model(
-        model_id="openlm-research/open_llama_3b_v2",
+        pretrained_model_name_or_path="google/gemma-7b",
         local_model_path=None,
     )  # load model gemma-2b-it from Hugging Face
     dialog_agent.load_tokenizer(
-        model_id="openlm-research/open_llama_3b_v2",
+        pretrained_model_name_or_path="google/gemma-7b",
         local_tokenizer_path=None,
     )  # load tokenizer for gemma-2b-it from Hugging Face
 
     # fine-tune loaded model with databricks-dolly-15k dataset
     # with default hyperparameters
-    # dialog_agent.fine_tune(data_path="databricks/databricks-dolly-15k")
+    # dialog_agent.fine_tune(data_path="GAIR/lima")
 
     # fine-tune loaded model with databricks-dolly-15k dataset
     # with customized hyperparameters
     # (`fine_tune_config` argument is optional. Defaults to None.)
-    dialog_agent.fine_tune(
-        "databricks/databricks-dolly-15k",
-        fine_tune_config={
-            "lora_config": {"r": 24, "lora_alpha": 48},
-            "training_args": {"max_steps": 300, "logging_steps": 3},
-        },
-    )
+    # dialog_agent.fine_tune(
+    #     "GAIR/lima",
+    #     fine_tune_config={
+    #         "lora_config": {"r": 24, "lora_alpha": 48},
+    #         "training_args": {"max_steps": 300, "logging_steps": 3},
+    #     },
+    # )
 
     user_agent = UserAgent()
 
diff --git a/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py b/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
index 10c5c583f..60a68ca31 100644
--- a/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
+++ b/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
@@ -56,15 +56,16 @@ def __init__(
 
     def load_model(
         self,
-        model_id: Optional[str] = None,
+        pretrained_model_name_or_path: Optional[str] = None,
         local_model_path: Optional[str] = None,
     ) -> None:
         """
         Load a new model into the agent.
 
         Arguments:
-            model_id (str): The Hugging Face model ID or a custom identifier.
-                            Needed if loading model from Hugging Face.
+            pretrained_model_name_or_path (str): The Hugging Face
+                             model ID or a custom identifier.
+                             Needed if loading model from Hugging Face.
             local_model_path (str, optional): Path to a locally saved model.
 
         Raises:
@@ -73,7 +74,10 @@ def load_model(
         """
 
         if hasattr(self.model, "load_model"):
-            self.model.load_model(model_id, local_model_path)
+            self.model.load_model(
+                pretrained_model_name_or_path,
+                local_model_path,
+            )
         else:
             logger.error(
                 "The model wrapper does not support dynamic model loading.",
@@ -81,14 +85,15 @@ def load_model(
 
     def load_tokenizer(
         self,
-        model_id: Optional[str] = None,
+        pretrained_model_name_or_path: Optional[str] = None,
         local_tokenizer_path: Optional[str] = None,
     ) -> None:
         """
         Load a new tokenizer for the agent.
 
         Arguments:
-            model_id (str): The Hugging Face model ID or a custom identifier.
+            pretrained_model_name_or_path (str): The Hugging Face model
+                            ID or a custom identifier.
                             Needed if loading tokenizer from Hugging Face.
             local_tokenizer_path (str, optional): Path to a locally saved
                                                   tokenizer.
@@ -99,13 +104,17 @@ def load_tokenizer(
         """
 
         if hasattr(self.model, "load_tokenizer"):
-            self.model.load_tokenizer(model_id, local_tokenizer_path)
+            self.model.load_tokenizer(
+                pretrained_model_name_or_path,
+                local_tokenizer_path,
+            )
         else:
             logger.error("The model wrapper does not support dynamic loading.")
 
     def fine_tune(
         self,
         data_path: Optional[str] = None,
+        output_dir: Optional[str] = None,
         fine_tune_config: Optional[Dict[str, Any]] = None,
     ) -> None:
         """
@@ -113,6 +122,11 @@ def fine_tune(
 
         Arguments:
             data_path (str): The path to the training data.
+            output_dir (str, optional): User specified path
+                                       to save the fine-tuned model
+                                       and its tokenizer. By default
+                                       save to this example's
+                                       directory if not specified.
 
         Raises:
             Exception: If fine-tuning fails or if the
@@ -120,7 +134,7 @@ def fine_tune(
         """
 
         if hasattr(self.model, "fine_tune"):
-            self.model.fine_tune(data_path, fine_tune_config)
+            self.model.fine_tune(data_path, output_dir, fine_tune_config)
             logger.info("Fine-tuning completed successfully.")
         else:
             logger.error("The model wrapper does not support fine-tuning.")
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index a7f86ef4b..692e690f6 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -10,7 +10,11 @@
 import os
 
 import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from transformers import (
+    AutoModelForCausalLM,
+    AutoTokenizer,
+    BitsAndBytesConfig,
+)
 from loguru import logger
 from dotenv import load_dotenv
 
@@ -31,9 +35,10 @@ class HuggingFaceWrapper(ModelWrapperBase):
     def __init__(
         self,
         config_name: str,
-        model_id: Optional[str] = None,
+        pretrained_model_name_or_path: Optional[str] = None,
         max_length: int = 512,
         data_path: Optional[str] = None,
+        output_dir: Optional[str] = None,
         device: Optional[torch.device] = None,
         local_model_path: Optional[str] = None,
         local_tokenizer_path: Optional[str] = None,
@@ -44,13 +49,19 @@ def __init__(
 
         Arguments:
             config_name (str): Configuration name for model setup.
-            model_id (str): Identifier for the pre-trained model on
-                            Hugging Face.
+            pretrained_model_name_or_path (str): Identifier for
+                                        the pre-trained model on
+                                        Hugging Face.
             max_length (int): Maximum sequence length for the
                               model output per reply.
                               Defaults to 512.
             data_path (str, optional): Path to the dataset for
                                        fine-tuning the model.
+            output_dir (str, optional): User specified path to save
+                                        the fine-tuned model
+                                        and its tokenizer. By default
+                                        save to this example's
+                                        directory if not specified.
             device (torch.device, optional): Device to run the model on.
                                              Default to GPU if available.
             local_model_path (str, optional): Local file path to a
@@ -62,7 +73,7 @@ def __init__(
         super().__init__(config_name=config_name)
         self.model = None
         self.max_length = max_length  # Set max_length as an attribute
-        self.model_id = model_id
+        self.pretrained_model_name_or_path = pretrained_model_name_or_path
         relative_path = os.path.join(
             os.path.dirname(__file__),
             "../conversation_with_agent_with_finetuned_model/.env",
@@ -78,9 +89,12 @@ def __init__(
         else:
             self.device = device
 
-        self.load_model(model_id, local_model_path=local_model_path)
+        self.load_model(
+            pretrained_model_name_or_path,
+            local_model_path=local_model_path,
+        )
         self.load_tokenizer(
-            model_id,
+            pretrained_model_name_or_path,
             local_tokenizer_path=local_tokenizer_path,
         )
 
@@ -89,6 +103,7 @@ def __init__(
                 self.model,
                 self.tokenizer,
                 data_path,
+                output_dir,
                 token=self.huggingface_token,
                 fine_tune_config=fine_tune_config,
             )
@@ -197,8 +212,9 @@ def format(
 
     def load_model(
         self,
-        model_id: Optional[str] = None,
+        pretrained_model_name_or_path: Optional[str] = None,
         local_model_path: Optional[str] = None,
+        fine_tune_config: Optional[Dict[str, Any]] = None,
     ) -> None:
         """
         Load a new model for the agent from
@@ -206,7 +222,8 @@ def load_model(
 
         Arguments:
             local_model_path (str): The file path to the model to be loaded.
-            model_id (str): An identifier for the model on Huggingface.
+            pretrained_model_name_or_path (str): An identifier for
+                                                 the model on Huggingface.
 
         Raises:
             Exception: If the model cannot be loaded from the given
@@ -216,25 +233,37 @@ def load_model(
                        or network issues while fetching the model.
         """
 
+        bnb_config_default = {}
+
+        if fine_tune_config is not None:
+            if fine_tune_config.get("bnb_config") is not None:
+                bnb_config_default.update(fine_tune_config["bnb_config"])
+
+        bnb_config = BitsAndBytesConfig(**bnb_config_default)
+
         try:
             if local_model_path is None:
                 self.model = AutoModelForCausalLM.from_pretrained(
-                    model_id,
+                    pretrained_model_name_or_path,
+                    quantization_config=bnb_config,
                     token=self.huggingface_token,
                     device_map="auto",
                 )
                 info_msg = (
-                    f"Successfully loaded new model '{model_id}' from "
+                    f"Successfully loaded new model "
+                    f"'{pretrained_model_name_or_path}' from "
                     f"Hugging Face"
                 )
             else:
                 self.model = AutoModelForCausalLM.from_pretrained(
                     local_model_path,
+                    quantization_config=bnb_config,
                     local_files_only=True,
                     device_map="auto",
                 )
                 info_msg = (
-                    f"Successfully loaded new model '{model_id}' from "
+                    f"Successfully loaded new model "
+                    f"'{pretrained_model_name_or_path}' from "
                     f"'{local_model_path}'"
                 )
 
@@ -245,7 +274,7 @@ def load_model(
             # Handle exceptions during model loading,
             # such as file not found or load errors
             error_msg = (
-                f"Failed to load model '{model_id}' "
+                f"Failed to load model '{pretrained_model_name_or_path}' "
                 f"from '{local_model_path}': {e}"
             )
 
@@ -255,7 +284,7 @@ def load_model(
 
     def load_tokenizer(
         self,
-        model_id: Optional[str] = None,
+        pretrained_model_name_or_path: Optional[str] = None,
         local_tokenizer_path: Optional[str] = None,
     ) -> None:
         """
@@ -264,8 +293,12 @@ def load_tokenizer(
         Arguments:
             local_tokenizer_path (str): The file path to the
                                         tokenizer to be loaded.
-            model_id (str): An identifier for the model on Huggingface.
-
+            pretrained_model_name_or_path (str): An identifier
+                                                for the model on Huggingface.
+            fine_tune_config (dict, optional): Configuration options for
+                                               fine-tuning the model,
+                                               including QLoRA and training
+                                               arguments.
         Raises:
             Exception: If the tokenizer cannot be loaded from the
             given path or identifier. Possible reasons include file not found,
@@ -275,13 +308,13 @@ def load_tokenizer(
         try:
             if local_tokenizer_path is None:
                 self.tokenizer = AutoTokenizer.from_pretrained(
-                    model_id,
+                    pretrained_model_name_or_path,
                     token=self.huggingface_token,
                 )
                 # log the successful tokenizer loading
                 logger.info(
                     f"Successfully loaded new tokenizer for model "
-                    f"'{model_id}' from Hugging Face",
+                    f"'{pretrained_model_name_or_path}' from Hugging Face",
                 )
 
             else:
@@ -291,15 +324,16 @@ def load_tokenizer(
                 # log the successful tokenizer loading
                 logger.info(
                     f"Successfully loaded new tokenizer for model "
-                    f"'{model_id}' from '{local_tokenizer_path}'",
+                    f"'{pretrained_model_name_or_path}'"
+                    f" from '{local_tokenizer_path}'",
                 )
-            self.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
 
         except Exception as e:
             # Handle exceptions during model loading,
             # such as file not found or load errors
             error_message = (
-                f"Failed to load tokenizer for model '{model_id}' from "
+                f"Failed to load tokenizer for model"
+                f" '{pretrained_model_name_or_path}' from "
                 f"'{local_tokenizer_path}': {e}"
             )
             logger.error(error_message)
@@ -309,6 +343,7 @@ def load_tokenizer(
     def fine_tune(
         self,
         data_path: Optional[str] = None,
+        output_dir: Optional[str] = None,
         fine_tune_config: Optional[Dict[str, Any]] = None,
     ) -> None:
         """
@@ -317,6 +352,11 @@ def fine_tune(
         Arguments:
             data_path (str): The file path to the training
                              data from Hugging Face.
+            output_dir (str, optional): User specified path
+                                       to save the fine-tuned model
+                                       and its tokenizer. By default
+                                       save to this example's
+                                       directory if not specified.
 
         Raises:
             Exception: If the fine-tuning process fails. This could be
@@ -328,6 +368,7 @@ def fine_tune(
                 self.model,
                 self.tokenizer,
                 data_path,
+                output_dir,
                 token=self.huggingface_token,
                 fine_tune_config=fine_tune_config,
             )
@@ -341,11 +382,71 @@ def fine_tune(
             )
             raise
 
+    def filer_sequence_lengths(
+        self,
+        max_input_seq_length: int,
+        dataset_obj: List[Dict[str, List[str]]],
+    ) -> List[int]:
+        """
+        Identifies and returns the indices of conversation
+          entries that exceed max_input_seq_length characters in length.
+
+        Args:
+            dataset_obj (List[Dict[str, List[str]]]): A list where
+                each dictionary contains 'conversations',
+                a list of two strings (question and answer).
+
+        Returns:
+            List[int]: Indices of conversations where the combined
+            length of the question and answer exceeds
+            max_input_seq_length characters.
+        """
+        # Initialize a list to store the sequence lengths
+        sequence_lengths = []
+
+        # list of indices that are too long
+        too_long = []
+
+        # Loop over the dataset and get the lengths of text sequences
+        for idx, example in enumerate(dataset_obj):
+            sequence_length = len(
+                example["conversations"][0] + example["conversations"][1],
+            )
+            sequence_lengths.append(sequence_length)
+            if sequence_length > max_input_seq_length:
+                too_long.append(idx)
+
+        return too_long
+
+    def formatting_prompts_func(
+        self,
+        example: Dict[str, List[List[str]]],
+    ) -> List[str]:
+        """
+        Formats each conversation in the dataset for training.
+        Args:
+            example (Dict[str, List[List[str]]]): A dataset.
+
+        Returns:
+            List[str]: A dataset with combined field.
+        """
+        output_texts = []
+        for i in range(len(example["conversations"])):
+            text = (
+                "### Question: "
+                + example["conversations"][i][0]
+                + "\n ### Answer: "
+                + example["conversations"][i][1]
+            )
+            output_texts.append(text)
+        return output_texts
+
     def fine_tune_training(
         self,
         model: AutoModelForCausalLM,
         tokenizer: AutoTokenizer,
         data_path: Optional[str] = None,
+        output_dir: Optional[str] = None,
         token: Optional[str] = None,
         fine_tune_config: Optional[Dict[str, Any]] = None,
     ) -> AutoModelForCausalLM:
@@ -361,10 +462,15 @@ def fine_tune_training(
                                        the pre-trained model.
             data_path (str): The file path or dataset identifier to load
                              the dataset from Hugging Face.
+            output_dir (str, optional): User specified path
+                                       to save the fine-tuned model
+                                       and its tokenizer. By default
+                                       save to this example's
+                                       directory if not specified.
             token (str): The authentication token for Hugging Face.
             fine_tune_config (dict, optional): Configuration options for
                                                fine-tuning the model,
-                                               including LoRA and training
+                                               including QLoRA and training
                                                arguments.
 
         Returns:
@@ -377,7 +483,7 @@ def fine_tune_training(
         Note:
             This method updates the model in place and also logs
             the fine-tuning process.
-            It utilizes the LoRA configuration and custom training arguments
+            It utilizes the QLoRA configuration and custom training arguments
             to adapt the pre-trained model to the specific dataset.
             The training log and trained model are saved in the same
             directory with the specific timestamp at saving time
@@ -387,8 +493,15 @@ def fine_tune_training(
         from datetime import datetime
         import json
 
-        dataset = load_dataset(data_path, token=token)
-        dataset = dataset["train"].train_test_split(test_size=0.1)
+        dataset = load_dataset(data_path, split="train", token=token)
+
+        indexes_to_drop = self.filer_sequence_lengths(300, dataset)
+
+        dataset_reduced = dataset.select(
+            i for i in range(len(dataset)) if i not in set(indexes_to_drop)
+        )
+
+        formatted_dataset = dataset_reduced.train_test_split(test_size=0.1)
 
         from peft import LoraConfig
 
@@ -406,70 +519,45 @@ def fine_tune_training(
 
         training_defaults = {
             "per_device_train_batch_size": 1,
-            # "gradient_accumulation_steps": 1,
+            "gradient_accumulation_steps": 4,
             "gradient_checkpointing": False,
-            # "max_steps": 10,
-            "num_train_epochs": 10,
+            "num_train_epochs": 5,
             "output_dir": "./",
             "optim": "paged_adamw_8bit",
-            "fp16": True,
             "logging_steps": 1,
-            "learning_rate": 1e-5,
-            # "num_train_epochs": 10.0,
         }
 
         if fine_tune_config is not None:
             if fine_tune_config.get("training_args") is not None:
                 training_defaults.update(fine_tune_config["training_args"])
 
+        if output_dir is not None:
+            training_defaults["output_dir"] = output_dir
+
         from peft import get_peft_model
 
         lora_config = LoraConfig(**lora_config_default)
         model = get_peft_model(model, lora_config)
 
-        from trl import SFTTrainer
         import transformers
+        from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
 
-        def formatting_func(example: Dict[str, Any]) -> Dict[str, str]:
-            if example.get("context", "") != "":
-                input_prompt = (
-                    f"Below is an instruction that describes a task, "
-                    f"paired with an input that provides further context. "
-                    f"Write a response that appropriately "
-                    f"completes the request.\n\n"
-                    f"### Instruction:\n"
-                    f"{example['instruction']}\n\n"
-                    f"### Input: \n"
-                    f"{example['context']}\n\n"
-                    f"### Response: \n"
-                    f"{example['response']}"
-                )
-            else:
-                input_prompt = (
-                    f"Below is an instruction that describes a task. "
-                    "Write a response that appropriately "
-                    f"completes the request.\n\n"
-                    "### Instruction:\n"
-                    f"{example['instruction']}\n\n"
-                    f"### Response:\n"
-                    f"{example['response']}"
-                )
-
-            return {"text": input_prompt}
-
-        formatted_dataset = dataset.map(formatting_func)
+        collator = DataCollatorForCompletionOnlyLM(
+            response_template=" ### Answer:",
+            tokenizer=tokenizer,
+        )
 
         trainer_args = transformers.TrainingArguments(**training_defaults)
 
         trainer = SFTTrainer(
             model,
+            formatting_func=self.formatting_prompts_func,
+            data_collator=collator,
             train_dataset=formatted_dataset["train"],
             eval_dataset=formatted_dataset["test"],
-            # data_collator=collator,
             peft_config=lora_config,
             args=trainer_args,
-            dataset_text_field="text",
-            max_seq_length=512,
+            max_seq_length=2048,
         )
 
         logger.info(
@@ -481,29 +569,49 @@ def formatting_func(example: Dict[str, Any]) -> Dict[str, str]:
         now = datetime.now()
         time_string = now.strftime("%Y-%m-%d_%H-%M-%S")
 
+        if output_dir is not None:
+            os.makedirs(output_dir, exist_ok=True)
+
         # Specify the filename
         log_name_temp = model.config.name_or_path.split("/")[-1]
         log_name = f"{log_name_temp}_{time_string}_log_history.json"
         log_path = os.path.join(os.path.dirname(__file__), log_name)
 
         # log training history
-        with open(log_path, "w", encoding="utf-8") as f:
-            json.dump(trainer.state.log_history, f)
+        if output_dir is not None:
+            with open(
+                os.path.join(output_dir, log_name),
+                "w",
+                encoding="utf-8",
+            ) as f:
+                json.dump(trainer.state.log_history, f)
+        else:
+            with open(log_path, "w", encoding="utf-8") as f:
+                json.dump(trainer.state.log_history, f)
 
         # save model
         model_name = (
             f"sft_{model.config.name_or_path.split('/')[-1]}_{time_string}"
         )
-        model_path = os.path.join(os.path.dirname(__file__), model_name)
+        if output_dir is not None:
+            model_path = os.path.join(output_dir, model_name)
+        else:
+            model_path = os.path.join(os.path.dirname(__file__), model_name)
         trainer.save_model(model_path)
 
         # save tokenizer
         tokenizer_name_temp = model.config.name_or_path.split("/")[-1]
         tokenizer_name = f"sft_{tokenizer_name_temp}_tokenizer_{time_string}"
-        tokenizer_path = os.path.join(
-            os.path.dirname(__file__),
-            tokenizer_name,
-        )
+        if output_dir is not None:
+            tokenizer_path = os.path.join(
+                output_dir,
+                tokenizer_name,
+            )
+        else:
+            tokenizer_path = os.path.join(
+                os.path.dirname(__file__),
+                tokenizer_name,
+            )
         tokenizer.save_pretrained(tokenizer_path)
 
         return model
diff --git a/examples/game_gomoku/code/board_agent.py b/examples/game_gomoku/code/board_agent.py
index 87247111b..f4c8641ea 100644
--- a/examples/game_gomoku/code/board_agent.py
+++ b/examples/game_gomoku/code/board_agent.py
@@ -25,7 +25,7 @@
 EMPTY_PIECE = "0"
 
 
-def board2img(board: np.ndarray, save_path: str) -> str:
+def board2img(board: np.ndarray, output_dir: str) -> str:
     """Convert the board to an image and save it to the specified path."""
 
     size = board.shape[0]
@@ -63,9 +63,9 @@ def board2img(board: np.ndarray, save_path: str) -> str:
     ax.set_xticklabels(range(size))
     ax.set_yticklabels(range(size))
     ax.invert_yaxis()
-    plt.savefig(save_path, bbox_inches="tight", pad_inches=0.1)
+    plt.savefig(output_dir, bbox_inches="tight", pad_inches=0.1)
     plt.close(fig)  # Close the figure to free memory
-    return save_path
+    return output_dir
 
 
 class BoardAgent(AgentBase):
diff --git a/examples/game_gomoku/main.ipynb b/examples/game_gomoku/main.ipynb
index df3149637..f7fc70f70 100644
--- a/examples/game_gomoku/main.ipynb
+++ b/examples/game_gomoku/main.ipynb
@@ -70,7 +70,7 @@
     "import matplotlib.pyplot as plt\n",
     "import matplotlib.patches as patches\n",
     "\n",
-    "def board2img(board: np.ndarray, save_path: str)->str:\n",
+    "def board2img(board: np.ndarray, output_dir: str)->str:\n",
     "    size = board.shape[0]\n",
     "    fig, ax = plt.subplots(figsize=(10, 10))\n",
     "    ax.set_xlim(0, size - 1)\n",
@@ -100,9 +100,9 @@
     "    ax.set_xticklabels(range(size))\n",
     "    ax.set_yticklabels(range(size))\n",
     "    ax.invert_yaxis()\n",
-    "    plt.savefig(save_path, bbox_inches='tight', pad_inches=0.1)\n",
+    "    plt.savefig(output_dir, bbox_inches='tight', pad_inches=0.1)\n",
     "    plt.close(fig)  # Close the figure to free memory\n",
-    "    return save_path"
+    "    return output_dir"
    ],
    "metadata": {
     "collapsed": false,
diff --git a/src/agentscope/web/README.md b/src/agentscope/web/README.md
index 19011ab4b..118e65dca 100644
--- a/src/agentscope/web/README.md
+++ b/src/agentscope/web/README.md
@@ -12,7 +12,7 @@ To start a web UI, you can run the following python code:
 import agentscope
 
 agentscope.web.init(
-    path_save="YOUR_SAVE_PATH",
+    path_save="YOUR_output_dir",
     host="YOUR_WEB_IP",         # defaults to 127.0.0.1
     port=5000                   # defaults to 5000
 )

From 9998e66e91ceaac5be68303dccf4f46655a1716e Mon Sep 17 00:00:00 2001
From: zyzhang <zyzhang1130@gmail.com>
Date: Wed, 8 May 2024 17:24:07 +0000
Subject: [PATCH 18/55] resolve issues raised

---
 ...rsation_with_agent_with_finetuned_model.py | 24 +++++++++----------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 9ec04c545..244b7ec5d 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -76,22 +76,22 @@ def main() -> None:
         model_config_name="my_custom_model",
     )
 
-    # (Optional) can load another model after
-    # the agent has been instantiated if needed
-    dialog_agent.load_model(
-        pretrained_model_name_or_path="google/gemma-7b",
-        local_model_path=None,
-    )  # load model gemma-2b-it from Hugging Face
-    dialog_agent.load_tokenizer(
-        pretrained_model_name_or_path="google/gemma-7b",
-        local_tokenizer_path=None,
-    )  # load tokenizer for gemma-2b-it from Hugging Face
+    # # (Optional) can load another model after
+    # # the agent has been instantiated if needed
+    # dialog_agent.load_model(
+    #     pretrained_model_name_or_path="google/gemma-7b",
+    #     local_model_path=None,
+    # )  # load model gemma-2b-it from Hugging Face
+    # dialog_agent.load_tokenizer(
+    #     pretrained_model_name_or_path="google/gemma-7b",
+    #     local_tokenizer_path=None,
+    # )  # load tokenizer for gemma-2b-it from Hugging Face
 
-    # fine-tune loaded model with databricks-dolly-15k dataset
+    # fine-tune loaded model with lima dataset
     # with default hyperparameters
     # dialog_agent.fine_tune(data_path="GAIR/lima")
 
-    # fine-tune loaded model with databricks-dolly-15k dataset
+    # fine-tune loaded model with lima dataset
     # with customized hyperparameters
     # (`fine_tune_config` argument is optional. Defaults to None.)
     # dialog_agent.fine_tune(

From f6b46ed25ee945f2658e8087dafa9c3ae0ff1031 Mon Sep 17 00:00:00 2001
From: zyzhang <zyzhang1130@gmail.com>
Date: Wed, 8 May 2024 17:31:29 +0000
Subject: [PATCH 19/55] resolve issues raised

---
 .../huggingface_model.py                              | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 692e690f6..1ce44c86c 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -8,6 +8,8 @@
 """
 from typing import Sequence, Any, Union, List, Optional, Dict
 import os
+from datetime import datetime
+import json
 
 import torch
 from transformers import (
@@ -15,6 +17,9 @@
     AutoTokenizer,
     BitsAndBytesConfig,
 )
+import transformers
+from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
+from datasets import load_dataset
 from loguru import logger
 from dotenv import load_dotenv
 
@@ -489,9 +494,6 @@ def fine_tune_training(
             directory with the specific timestamp at saving time
             as part of the log/model fodler name.
         """
-        from datasets import load_dataset
-        from datetime import datetime
-        import json
 
         dataset = load_dataset(data_path, split="train", token=token)
 
@@ -539,9 +541,6 @@ def fine_tune_training(
         lora_config = LoraConfig(**lora_config_default)
         model = get_peft_model(model, lora_config)
 
-        import transformers
-        from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
-
         collator = DataCollatorForCompletionOnlyLM(
             response_template=" ### Answer:",
             tokenizer=tokenizer,

From 6bf09f1ca9a4f07d1a5ca01d631da7e46c116a58 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Fri, 10 May 2024 17:30:50 +0800
Subject: [PATCH 20/55] Update README.md

updated the dependencies needed
---
 examples/conversation_with_agent_with_finetuned_model/README.md | 1 +
 1 file changed, 1 insertion(+)

diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
index 366f2842c..a1ad7fe45 100644
--- a/examples/conversation_with_agent_with_finetuned_model/README.md
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -67,6 +67,7 @@ Before running this example, ensure you have installed the following packages:
 - `python-dotenv`
 - `datasets`
 - `trl`
+- `bitsandbytes`
 
 Additionally, set `HUGGINGFACE_TOKEN` in the `agentscope/examples/conversation_with_agent_with_finetuned_model/.env`.
 

From 8d7e8808aec4ac3f97a73bae4fea3d48df2503d9 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Fri, 10 May 2024 17:51:49 +0800
Subject: [PATCH 21/55] Update README.md

---
 examples/conversation_with_agent_with_finetuned_model/README.md | 1 -
 1 file changed, 1 deletion(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
index a1ad7fe45..d9f682a5f 100644
--- a/examples/conversation_with_agent_with_finetuned_model/README.md
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -63,7 +63,6 @@ The example is tested using specific Hugging Face model `google/gemma-7b` on dat
 Before running this example, ensure you have installed the following packages:
 
 - `transformers`
-- `peft`
 - `python-dotenv`
 - `datasets`
 - `trl`

From 98b471ee164736ec145d28d9836aadbba86c7a5c Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Mon, 20 May 2024 22:37:14 +0800
Subject: [PATCH 22/55] Update huggingface_model.py

Updated the way to read token from .env file, so that it can work in any example directory.
---
 .../huggingface_model.py                                  | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 1ce44c86c..c5c2cbf3e 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -79,11 +79,9 @@ def __init__(
         self.model = None
         self.max_length = max_length  # Set max_length as an attribute
         self.pretrained_model_name_or_path = pretrained_model_name_or_path
-        relative_path = os.path.join(
-            os.path.dirname(__file__),
-            "../conversation_with_agent_with_finetuned_model/.env",
-        )
-        dotenv_path = os.path.normpath(relative_path)
+        script_path = os.path.abspath(__file__)
+        script_dir = os.path.dirname(script_path)
+        dotenv_path = os.path.join(script_dir, ".env")
         _ = load_dotenv(dotenv_path)  # read local .env file
         self.huggingface_token = os.getenv("HUGGINGFACE_TOKEN")
 

From ee062fd4a256ab7c08e2c4a4f59493248bf13df8 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 22 May 2024 20:23:44 +0800
Subject: [PATCH 23/55] reverted unnecessary changes

---
 src/agentscope/web/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/agentscope/web/README.md b/src/agentscope/web/README.md
index 855e2d831..1fa92ce7d 100644
--- a/src/agentscope/web/README.md
+++ b/src/agentscope/web/README.md
@@ -12,7 +12,7 @@ To start a web UI, you can run the following python code:
 import agentscope
 
 agentscope.web.init(
-    path_save="YOUR_output_dir",
+    path_save="YOUR_SAVE_PATH",
     host="YOUR_WEB_IP",         # defaults to 127.0.0.1
     port=5000                   # defaults to 5000
 )

From 6203f511829d9365bc2d16d05b232a7b8526c677 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 22 May 2024 20:29:27 +0800
Subject: [PATCH 24/55] Revert back unnecessary changes

---
 docs/sphinx_doc/en/source/tutorial/105-logging.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/sphinx_doc/en/source/tutorial/105-logging.md b/docs/sphinx_doc/en/source/tutorial/105-logging.md
index 7575b11cd..98f872a8b 100644
--- a/docs/sphinx_doc/en/source/tutorial/105-logging.md
+++ b/docs/sphinx_doc/en/source/tutorial/105-logging.md
@@ -75,7 +75,7 @@ You can run the WebUI in the following python code:
 import agentscope
 
 agentscope.web.init(
-    path_save="YOUR_output_dir"
+    path_save="YOUR_SAVE_PATH"
 )
 ```
 

From ce0671f01e1dec999eeb40470fd01b64b78b44e5 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 22 May 2024 20:32:09 +0800
Subject: [PATCH 25/55] revert unnecessary change

---
 docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md b/docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md
index 073d30515..d517cd46b 100644
--- a/docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md
+++ b/docs/sphinx_doc/zh_CN/source/tutorial/105-logging.md
@@ -76,7 +76,7 @@ logger.error("The agent encountered an unexpected error while processing a reque
 import agentscope
 
 agentscope.web.init(
-    path_save="YOUR_output_dir"
+    path_save="YOUR_SAVE_PATH"
 )
 ```
 

From 456f1aeeb9eb314fee8e1ef4555d17015bd6a60f Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 22 May 2024 20:36:26 +0800
Subject: [PATCH 26/55] revert back unnecessary changes

---
 examples/game_gomoku/code/board_agent.py |   6 +-
 examples/game_gomoku/main.ipynb          | 260 ++++++++++++-----------
 2 files changed, 143 insertions(+), 123 deletions(-)

diff --git a/examples/game_gomoku/code/board_agent.py b/examples/game_gomoku/code/board_agent.py
index d84051277..6cbef4ced 100644
--- a/examples/game_gomoku/code/board_agent.py
+++ b/examples/game_gomoku/code/board_agent.py
@@ -25,7 +25,7 @@
 EMPTY_PIECE = "0"
 
 
-def board2img(board: np.ndarray, output_dir: str) -> str:
+def board2img(board: np.ndarray, save_path: str) -> str:
     """Convert the board to an image and save it to the specified path."""
 
     size = board.shape[0]
@@ -63,9 +63,9 @@ def board2img(board: np.ndarray, output_dir: str) -> str:
     ax.set_xticklabels(range(size))
     ax.set_yticklabels(range(size))
     ax.invert_yaxis()
-    plt.savefig(output_dir, bbox_inches="tight", pad_inches=0.1)
+    plt.savefig(save_path, bbox_inches="tight", pad_inches=0.1)
     plt.close(fig)  # Close the figure to free memory
-    return output_dir
+    return save_path
 
 
 class BoardAgent(AgentBase):
diff --git a/examples/game_gomoku/main.ipynb b/examples/game_gomoku/main.ipynb
index 7044a8243..569321c15 100644
--- a/examples/game_gomoku/main.ipynb
+++ b/examples/game_gomoku/main.ipynb
@@ -31,6 +31,13 @@
   {
    "cell_type": "code",
    "execution_count": 1,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:57.602994Z",
+     "start_time": "2024-03-27T08:09:57.565787Z"
+    },
+    "collapsed": false
+   },
    "outputs": [],
    "source": [
     "YOUR_MODEL_CONFIGURATION_NAME = \"{YOUR_MODEL_CONFIGURATION_NAME}\"\n",
@@ -39,38 +46,38 @@
     "    \n",
     "    # ...\n",
     "}"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:57.602994Z",
-     "start_time": "2024-03-27T08:09:57.565787Z"
-    }
-   }
+   ]
   },
   {
    "cell_type": "markdown",
+   "metadata": {
+    "collapsed": false
+   },
    "source": [
     "## Step 1: Prepare a board for Gomoku\n",
     "\n",
     "First we create a `BoardAgent` class by inheriting from `AgentBase`, which manage the game board and update the game status as follows.\n",
     "\n",
     "To create a better visual experience, we also provide a function `board2img` to convert the board to an image. "
-   ],
-   "metadata": {
-    "collapsed": false
-   }
+   ]
   },
   {
    "cell_type": "code",
    "execution_count": 2,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:58.153932Z",
+     "start_time": "2024-03-27T08:09:57.571266Z"
+    },
+    "collapsed": false
+   },
    "outputs": [],
    "source": [
     "import numpy as np\n",
     "import matplotlib.pyplot as plt\n",
     "import matplotlib.patches as patches\n",
     "\n",
-    "def board2img(board: np.ndarray, output_dir: str)->str:\n",
+    "def board2img(board: np.ndarray, save_path: str)->str:\n",
     "    size = board.shape[0]\n",
     "    fig, ax = plt.subplots(figsize=(10, 10))\n",
     "    ax.set_xlim(0, size - 1)\n",
@@ -100,29 +107,30 @@
     "    ax.set_xticklabels(range(size))\n",
     "    ax.set_yticklabels(range(size))\n",
     "    ax.invert_yaxis()\n",
-    "    plt.savefig(output_dir, bbox_inches='tight', pad_inches=0.1)\n",
+    "    plt.savefig(save_path, bbox_inches='tight', pad_inches=0.1)\n",
     "    plt.close(fig)  # Close the figure to free memory\n",
-    "    return output_dir"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:58.153932Z",
-     "start_time": "2024-03-27T08:09:57.571266Z"
-    }
-   }
+    "    return save_path"
+   ]
   },
   {
    "cell_type": "markdown",
-   "source": [
-    "The following code shows the implementation of the `BoardAgent` class. The agent manages the game board and updates the game status."
-   ],
    "metadata": {
     "collapsed": false
-   }
+   },
+   "source": [
+    "The following code shows the implementation of the `BoardAgent` class. The agent manages the game board and updates the game status."
+   ]
   },
   {
    "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.392069Z",
+     "start_time": "2024-03-27T08:09:58.159470Z"
+    },
+    "collapsed": false
+   },
    "outputs": [],
    "source": [
     "import numpy as np\n",
@@ -229,18 +237,13 @@
     "            else:\n",
     "                count = 0\n",
     "        return False\n"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.392069Z",
-     "start_time": "2024-03-27T08:09:58.159470Z"
-    }
-   },
-   "execution_count": 3
+   ]
   },
   {
    "cell_type": "markdown",
+   "metadata": {
+    "collapsed": false
+   },
    "source": [
     "### Step2: Prepare a Gomoku Player Agent\n",
     "\n",
@@ -250,13 +253,18 @@
     "2. Within the agent, to enable the agent to think, we ask LLMs to respond in a dictionary format, which contains the thought and the move (\"thought\" field must come before \"move\"). To achieve this, we prepare a parsing function to extract the dictionary from response. \n",
     "\n",
     "The implementation of the Gomoku agent is as follows: "
-   ],
-   "metadata": {
-    "collapsed": false
-   }
+   ]
   },
   {
    "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.397713Z",
+     "start_time": "2024-03-27T08:09:59.392729Z"
+    },
+    "collapsed": false
+   },
    "outputs": [],
    "source": [
     "import json\n",
@@ -327,15 +335,7 @@
     "        \n",
     "        # Hide thought from the response\n",
     "        return Msg(self.name, response[\"move\"], role=\"assistant\") \n"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.397713Z",
-     "start_time": "2024-03-27T08:09:59.392729Z"
-    }
-   },
-   "execution_count": 4
+   ]
   },
   {
    "cell_type": "markdown",
@@ -346,17 +346,25 @@
   },
   {
    "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.446367Z",
+     "start_time": "2024-03-27T08:09:59.398116Z"
+    },
+    "collapsed": false
+   },
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m2024-03-27 16:09:59.427\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models\u001B[0m:\u001B[36mread_model_configs\u001B[0m:\u001B[36m171\u001B[0m - \u001B[1mLoad configs for model wrapper: post_api\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.435\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m_create_monitor_table\u001B[0m:\u001B[36m341\u001B[0m - \u001B[1mInit [monitor_metrics] as the monitor table\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.435\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m_create_monitor_table\u001B[0m:\u001B[36m342\u001B[0m - \u001B[1mInit [monitor_metrics_quota_exceeded] as the monitor trigger\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.436\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m311\u001B[0m - \u001B[1mSqliteMonitor initialization completed at [./runs/run_20240327-160958_kbf2ta/agentscope.db]\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.440\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models.model\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m257\u001B[0m - \u001B[1mInitialize model [post_api]\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.441\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models.model\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m257\u001B[0m - \u001B[1mInitialize model [post_api]\u001B[0m\n"
+      "\u001b[32m2024-03-27 16:09:59.427\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models\u001b[0m:\u001b[36mread_model_configs\u001b[0m:\u001b[36m171\u001b[0m - \u001b[1mLoad configs for model wrapper: post_api\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.435\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m_create_monitor_table\u001b[0m:\u001b[36m341\u001b[0m - \u001b[1mInit [monitor_metrics] as the monitor table\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.435\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m_create_monitor_table\u001b[0m:\u001b[36m342\u001b[0m - \u001b[1mInit [monitor_metrics_quota_exceeded] as the monitor trigger\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.436\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m311\u001b[0m - \u001b[1mSqliteMonitor initialization completed at [./runs/run_20240327-160958_kbf2ta/agentscope.db]\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.440\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models.model\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m257\u001b[0m - \u001b[1mInitialize model [post_api]\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.441\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models.model\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m257\u001b[0m - \u001b[1mInitialize model [post_api]\u001b[0m\n"
      ]
     }
    ],
@@ -381,18 +389,13 @@
     ")\n",
     "\n",
     "board = BoardAgent(name=\"Host\")"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.446367Z",
-     "start_time": "2024-03-27T08:09:59.398116Z"
-    }
-   },
-   "execution_count": 5
+   ]
   },
   {
    "cell_type": "markdown",
+   "metadata": {
+    "collapsed": false
+   },
    "source": [
     "### Step 3: Start the Gomoku game\n",
     "\n",
@@ -401,26 +404,32 @@
     "In this game, we use a message hub to share messages between the two players (board agent doesn't have memory), so that one player can hear what the board agent says to the other player, and what moves the other player makes. \n",
     "\n",
     "Note here we only show 10 steps of the game. You can adjust the `MAX_STEPS` to play more rounds."
-   ],
-   "metadata": {
-    "collapsed": false
-   }
+   ]
   },
   {
    "cell_type": "code",
    "execution_count": 6,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:10:33.989202Z",
+     "start_time": "2024-03-27T08:09:59.446824Z"
+    },
+    "collapsed": false
+   },
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: Welcome to the Gomoku game! Black player goes first. Please make your move.\n"
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: Welcome to the Gomoku game! Black player goes first. Please make your move.\n"
      ]
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAe5UlEQVR4nO2da5BU1bn3n75Nd0/3DDPcQQQEBggiHiEKqOAxr0RS1ouHypATIppLqbGiRYwfktIylUvpm1SSDzExSSXlJQmGBC9RbqKojCgwDMMAM8PQMMPcunu6Z/q6u/fuvXvv3nut9wN0XuvUyWuvpxnm7Mrzq5qyLPq31+r99L8vu9fqx8E5B4Ig7IVzoidAEIQ4FFyCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAhwsE9d+5cw44dO9aOjo7WjseECIL4dByiCzC2bt26obGx0Zw9e7b69NNPt+m67uacu8ZpfgTxL4Pb7TbdbrdV0W1FD26apnvp0qWJRCIxSdd15969ez97+vTpZ5YvXz5L9FiMMTh79iysWLFCVAXOOXR1dcGNN94o7AIAdHZ2olxFUWBsbAwWLlwo7GqaBpFIBBYvXizs6roOAwMD8JnPfEbYNU0TLly4ANdff72wyxiDnp4euOGGG4Rdzjl0d3df9frm83lIpVKwYMECYbdQKEAsFoOmpiZht1gswtDQECxdulTYLZVK3O/3/58tW7bsrEjgnAv9vfrqq/MffPDBL7W3t0/jnIOu655XXnmliyMolUr8z3/+M0bljDH+xz/+Ee2+/PLLKDcWi/EDBw6g3FQqxXfv3o1y8/k8f+2111Cuqqp8586dKNcwDL5jxw6Ua1kW/9Of/oRyq6lRJBLhBw8eRLljY2N83759KDebzfK///3vKFdRFHPXrl2P8ApzKPyKu2XLlqEtW7YMiXoEQVw56KoyQdiQCQ2uw+GYyOEJwrZMaHA5bSkkCBT0VpkgbAgFlyBsCAWXIGwIBZcgbAgFlyBsCAWXIGwIfY9LEDZEOLhHjhy59umnn77vSgxO3+MSBA7htcrz5s2Lh0IhxhhzMMbg3Llzc8LhsPfMmTPCgzPGIBwOA8blnKNdAIBIJIJyU6kUDAwMoNxcLgeDg4MoV1VVGBoaQrmGYcDw8DDKtSwL7U5UjcbGxtDnKpvNouurKAq6vsViEXRdr/j2wsF1OBweXdfrc7mcr66ursgYc3DOgTEmeihgjP3jDwN23E+OjRkTO241LmPMdm4197fs/yudZxGEg+t0On0PP/xwbzqdrm1sbNSWL18eCYVC+sqVK0UPBaZpQigUAozLL+/1xLpdXV0oNx6Pg2VZKDedTkM+n0e5sizD6OgoytU0DQYHB1FuqVSCCxcuoNzyXt6rXaNoNAoulwvlJhIJKBaLKFeSJEin0yi3UChANBqt+PbCwZ09e3Z206ZNLaIeQRBXDvo6iCBsCAWXIGwIBZcgbAgFlyBsCK2cIggbQr+AQRA2hN4qE4QNoeAShA2h4BKEDREObjab9SYSiUmmadKVJYKYIISXPL7wwguzQqHQzevXrz/zta99rW88JkUQxP8f4eA+8MADkccee+z6+vp6tdz06/jx40GXS7xhH2MMjh8/Dh6PR9jlnENbWxt4vV5hFwCgra0NfD6fsCdJEgwPD4MkScKuoihw4cIFUFVV2NU0Dbq6usA0TWHXMAw4efIk6us3y7KgtbUV3G7hh8o/6ltTUyPsluuLqVE6nYZYLAbpdFrYzefz0N/fD7IsC7uqqkJPT4/Q9rwyuq4L1Va4Gs8+++wtfr9//rJly1q9Xi/7/Oc/36koSuHee+8VPRRYlgWapgHG5ZyDoigoF+DSbhuMOzo6Ct3d3bBhwwZhN5PJQENDA9xzzz3CrizL4PV6UXMuFosAACjXNE3QdR3lMsZAVdWrXt+RkRG4cOECfO5znxN2k8kkdHR0wMaNG4XdXC4HgUAANm3aJOyqqgrvvfdexbcXDu6PfvSjM4ZhXAgEAjIAgM/nK3k8Hu73+0UPBaZpQk1NDWBcznlVrsfjQbk+nw89bjVuNecKAND3t1QqocdljP1L1UjX9arOldNZ+SUn4eA2NjZqAKCJegRBXDno6yCCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAhwsE9evTo1H379q1KpVLiq78JgrgiCC95HBgY0F988cXFX/rSl4rf+ta3ekzTdDHGULtWLMsCy7JQLucc7QJcWht6tedcjWuaZlXjYu9vNeMyxqqqbzVztmN9RX6DTTi4a9as0drb2xOqqnoub+tb1draWid6HID/t+0LsyUQ4NLWPMyWQABAbyfM5XIQDochk8kIu4VCAXp7e0FRFGG3WCxCd3c3GIYh7JZKJejo6EA3ozp+/LjQAvgynHM4fvw4aksgAL6+2WwWYrEYJBIJYVeWZRgYGIBcLifsapoG586dA00TX8ovWlfhM7p3797rXC5X49q1a0Ner5dt3rz5hGma8n33ibfMNU0TnE4n3H///cJu+RkZ61qWhXLj8Th0dnaitn2l02k4evQoatuXLMvQ2NgIzc3Nwq6maeD3+2Hr1q3CbqlUApfLBdu2bRN2y08UV7tG0WgUQqEQautlIpGA9vZ21NZLSZKgpaUFNm/eLOwWCgXYv39/xbcXDu4TTzzRBwBX5Jcv6HeVCQIH/a4yQdgQ+jqIIGwIBZcgbAj1DiIIG0KfcQnChtBbZYKwIRRcgrAhFFyCsCEUXIKwIXRVmSBsiHBwc7mc8/HHH//f0Wh0SrWD01VlgsAhvFZ59+7di/7617/OffDBBztnzZqV6e/vnxGPxz3nz58XHpwxBvF4HDAu5xztAlzqAYRxU6kURKNRlJvL5WBkZATlFgoFiMViKFfXdfS5sizLdjUaGxtD1yiTyaBdWZbR9dU0TWiHkHBwJUmas2jRommDg4OzlixZEpEkKVAoFFyYbW6MMVAUBbVFDuDSgxnrYseVJAlkWUa5+Xwe7Wqahp6zYRho17IstMs5r6pGWLeaGlXjFgoF9LkqFotC2y6Fg7t9+/ZDN998c8/SpUsVj8fDV61aNTAwMFC89dZbRQ8FpmnC4OAgYFzOOfT19aHd3t5elBuPx8HtdqPcdDoNhmGgXFmWIZvNolxN0yAej6PcUqkEw8PDKJcxBv39/Ve9RtFoFGpra1FuIpEAzjnKlSQJFEVBuYVCAVKpVMW3R+1wXrt27RjGIwjiykBfBxGEDaGvgwjChtAmA4KwIfRWmSBsCAWXIGwIBZcgbAgFlyBsCF1VJggbQleVCcKGCAd3eHh41oEDB9YoiuIfjwkRBPHpCC95/Oijj65vb2/3rV27lgMAXG765SiVSsKDl5skYdxyiwqMWx4b45abYJH76ZSbfmFrxBiz1f2t1h3Xpl9z5sw5feTIkZUvvvjisscee+zM5aZfQczn1XJDKUxTKM45tLW1QU1NjbALAHDixAnwer3CniRJEA6HIZvNCruKokBvby8UCgVht1gsQldXF+pBUSqV4OTJk8IewKUnOGxjtnKNsI3ZsPXNZDIQi8WEFu2XyefzMDAwAPl8XthVVRXOnTsHxWJR2NV1Xej2wok5c+ZMQyAQmLR27dqeK9H0y+FwoBpKlZt+YV3TNFHuRDb9amhoQDf98vl8E9L0i3N+1WtETb/+G77zne/0A0C/qEcQxJWDvsclCBtCwSUIG0LBJQgbQsElCBtCwSUIG0LBJQgbQsElCBtCwSUIG0LBJQgbQsElCBsiHNwPP/xw2ve+971Nb7755vTxmBBBEJ+O8FrlX/3qV2scDof/pptucliW5bh48eLMWCxWEwqFhAe3LAtisRhgXM452gW4tFkA4yaTSQiHwyhXkiSIRCIot1Ao/GPxvCi6rsPIyAjKNU0TfZ4ZY1XVF1uj0dFRdI0ymQy6RrIso2s07k2/5s6dqwNA8ezZs9c3NzcnZFn2a5rmzOVyoocCxhioqgoYl3MOmqahXABAu7Iso+dcjauqKnrOhmGgXcuyqEYVoigK2r0aTb9aT548eX1TU1PI7Xbzm266afDixYvFNWvWiB4KTNOE/v5+wLjlplBY9/z58yg3Ho+Dy+VCuel0GorFIsotd5DDuJqmwcjICMotlUowODiIchlj0NfXd9VrFI1Gwe/3o9xEIgGWZaHccqc/jFsoFCCRSFR8e+HgLliwQF6wYMFxUY8giCsHXVUmCBtCwSUIG0LBJQgbQsElCBtCnQwIwoZQJwOCsCH0VpkgbAgFlyBsCAWXIGyI8MopWZY9pmn6fD5f0e/345rCEARRFcKvuF1dXTMeeeSR//zLX/6yfDwmRBDEpyP8invbbbdFf/azn0nr1q3r13XdeeDAgZtOnDgRxDTQYoxBe3s7+P3iHTs559De3g6BQEDYBQA4efIkBINBYU+SJBgaGgJFUYRdRVHg/PnzQtu3ymiaBp2dncIewKWNAidOnEA137IsC06cOAE+n0/YLdeotrZW2AXA1yidTsPIyAhql04+n4eLFy+CpmnCrqqqcPbsWbAsS9jVdV2o8ZdwcHt7e4M+n89csmRJHgBg/fr1oXQ6rd59992ihwLLsiCXywHG5ZxDJpNBuQAAqVQK5Y6NjUF3dzfcddddwm4mkwG/348at/xEgXGLxSIUi0WUa5omyLKMchljkM1m0fXF1mhkZAR6e3vhzjvvFHZTqRTU19ejxs3lcuB2u1GuqqrQ0tJS8e2Fg3vNNdeov/71r98p/38wGCz6fD5WV1cneigwTRN8Ph9gXM75hLiKooDf70e5hmGgXQBAu263G31/S6US2mWMTUiNgsEg+lxpmoZ2LctCu06nE5zOyj+5Cgc3EAiwQCAg3gD0v4FWThEEDlo5RRA2hNYqE4QNoVdcgrAhtHKKIGwIBZcgbAgFlyBsCAWXIGwIBZcgbAgFlyBsiHBww+Fw/fvvv39jb28vbuX4J6DvcQkCh/CSxx/84Ac3Hz58ePEjjzzi/O53v3uaMebgnAv1PSnDGAOsyzlHu2X/ao9bjTtR56qace0452rryxiras6VIhxcy7L4lClT0m6326vruvPNN9+85ejRo3XYyba2tgp7AJdO0rFjx4QWZn+SY8eOgcvlEvYkSYJwOAzJZFLYVRQF+vr60E2huru7UdvNDMOAU6dOgWmawq5lWdDW1oZaLMMYg2PHjqHeWXHOobW1FVWjTCYD8XgcRkdHhd18Pg8DAwOQyWSEXVVV4dy5c1AoFIRdXdfHd5PBunXrMoqizGxsbEx4vV72xS9+sY0xJt93332ihwLTNMHlcsH9998v7JYfSA888ADKZYyh3Hg8Dp2dnbBx40ZhN51Ow9GjR2HTpk3CrizLMHnyZGhubhZ2NU2DQCAAW7duFXZLpRJ4PB7Ytm2bsMsYA6fTia4R5xzllltdbtiwQdhNJBLQ3t4O99xzj7ArSRK0tLTA5s2bhd1CoQD79++v+PbCwX3ooYfOPPTQQ2dEPYIgrhx0VZkgbAgFlyBsCAWXIGwIBZcgbAgFlyBsCAWXIGwIBZcgbAgFlyBsCAWXIGwI/cojQdgQ4eDu3bt34fbt2zd1dXU1Vjs4/cojQeAQXqu8c+fOpmg0yl999dWVK1as+CAcDk9NJpOegYEB4cEty4JEIgEYl3MOyWQS5QIA2k0mkzA6OopyJUmCsbExlFsoFNDnStd1tGuapu1qNDo6iq5RJpNBu7Iso+uraRqUSpV3rRUO7vbt20/+7ne/+5zH40mWSiVHLBZrTKfT7kgkInooYIxBJpMBjFtu+oVxAQDtZjIZSCaTKDefz6NdVVUhlUqhXMMwIJ1Oo1zLsqqqEXZcAHyNkskk+jxLkoQ+z4qioN1isSi07VI4uFOnTnXdeeedsQ0bNpzxeDx8zZo1fZFIRLvjjjtEDwWmaUI0GgWMyzmH4eFhtDs4OIhy4/E4+P1+lJtOp4FzjnJlWYZCoYByNU2DVCqFckulEsRiMZTLGINwOHzVaxSNRmHSpEkoN5FIgNvtRrmSJIFhGCh33Lf1NTU1jTU1NY2JegRBXDno6yCCsCEUXIKwIfQ9LkHYEOrWRxA2hF5xCcKG0CsuQdgQujhFEDaEgksQNoQ+4xKEDakouLquu1paWm4cGxsLvP/++0vb2tpmXInB6TMuQeCoaMmjZVnwk5/8ZNGjjz4a2LVr13Rd12tee+2115xOJ2eMOTnnYFmW8ODlBkkYF+BS8K+2W82cq3WrmfNE3N9yq5eJqJFd61spFQW3trbWWrhwYRwAfLfccktqeHj4GgCAy02/bq626Re2KVQ1Tb+wDaVyuRwMDw9DKpUSdstNv/L5vLBbTdOvUqkEHR0dVTX9wjBRNaq26dfg4CBks1lhV1VVCIVC/3OafhWLRU8gEFh49uzZ3ODgoG/mzJkxp9PJvV4vp6ZflTORTb9qa2snpOmXw+G46jWipl+X8fl8pV/84hc7hGdDEMS4QFeVCcKG0MopgrAhtACDIGwIBZcgbAgFlyBsCAWXIGwIBZcgbAgFlyBsCAWXIGwIBZcgbEhFwZVl2fPkk09u6ezsnP3tb3/7i0eOHJk23hMjCOKfU9Fa5bq6ulIkEklwzn2zZs1KnT9/Pnj77bcnAQCGh4enJRIJT39/v/Dg5aZfGJdzjnYBLvWXwbjJZBLi8TjKlSQJRkdHUW6hUICxsTGUW276hXHLTb8wLmNsQmo0OjqKrlF5ZxHGlWUZXd9xa/rl9Xo9xWLR19DQ4DEMwwcAUCqVHKOjo5Oy2ax7ZGREeLKMMchms4BxOedoFwDQbiaTgXQ6jXLz+TzaVVUVMpkMyjUMA+2Wm37ZqUbJZBJSqRTKlSQJXSNFUdDnalyafum67t6wYQMbGRkJNjQ0sMmTJwcAADweD1+9evXFcDisrV+/XniypmlCJBIBjMs5h6GhIbQ7MDCAcuPxOPh8PpSbTqeBMYZyZVkGRVFQrqZpkEwmUW6pVIKRkRGUyxiD4eHhq16jaDQK9fX1KDeRSIDL5UK5kiSBrusod1y29Xm9XvPLX/7yIeHZEAQxLtBVZYKwIRRcgrAhtJGeIGwIbaQnCBtCb5UJwoZQcAnChlBwCcKGUHAJwoZQcAnChlQc3EKh4DNN06koik9VVfG+EARBXDEqCm4ul6tpbm7eduzYsaYnn3zy37dv3/758Z4YQRD/nIrWKk+aNMmYO3duaOrUqUZzc3Po+eefXwYAYBiG8+DBgytOnToVCAaDwoMzxuDUqVNQX18v7HLO4dSpU9DQ0CDsAgCcPn0aGhsbhb1sNguDg4Og67qwK8syhEIh1PfXmqbB6dOnwePxCLuGYUBHRwfU1tYKu5ZlQUdHB9TV1Qm75RpNmjRJ2AXA1yidTkM0GgVVVYXdXC4HfX19qAZphUIBuru7UU3OdF0XahZWUXANw3ApijL3ww8/rHvppZdWPPbYY/sBAGpqatgtt9zSH4vFtHXr1glP1rIsGBsbA4zLOYdYLIZyAS7tIMG4iUQCAoEAys1ms+B0OlGuoihQLBZRrq7rkMvlUK5pmpBMJlEuYwzi8Ti6RiMjIyg3FotBY2Mjyk2lUlBTU4Ny8/k8WJaFclVVhY8//rji21cUXJfLxZ599tnDlmXxu+++u8fn8/3jqaGhoaEQCATY5MmThSdrmiYEAgHAuJzzCXF1XYe6ujr0uMFgEOV6PB60q2ka+v6WSiW0yxirqka1tbUoV1VV9LkyTRNdX6fTiR7X6/UKtRStNLh8/vz5MeHZEAQxLtDXQQRhQyi4BGFDKLgEYUMouARhQyi4BGFDKLgEYUMouARhQyi4BGFDKLgEYUMqCq5hGM62tralqVQq2NbWtqSzs3PqlRicfuWRIHBUtOSxVCo5nnrqqRU//OEPrfPnzzfs27fvpt27d/8NAIBz7rj8X+HBOef/+MO42HH/6zEw3tUe1473t5r6fvIYWGci6lvt/a2UioIbCASsxYsXRxsbG82mpqa0w+GYDwCg67rzjTfeWP3xxx/XYbZBMcagtbUVXZyjR48Ke590Ma/4kiRBJBKBRCIh7CqKAn19fZDNZoXdYrEI3d3dQlu/yhiGAadOnQLDMIRdy7Kgra0NGGPCbrm+GKqpUbnjXiwmvrw+n8/D4OAgpNNpYVdVVQiFQiDLsrCr67rQdsCKglssFj1ut7tpz549051Op7Vo0aI0AIDX62XNzc3HOefyfffdJzxZ0zTB7XbD/fffL+xyzsHhcMBXv/pVlAsAKDcej0NnZyds3LhR2E2n03D06FHYtGmTsCvLMrz77rvQ3Nws7GqaBsFgELZu3SrslkolqKmpgW3btgm7jDFwuVzwwAMPCLvV1CgajUIoFIINGzYIu4lEAtrb2+Gee+4RdiVJgpaWFti8ebOwOy5Nv3w+X+m55577c/lkulwu+iVzgphAKu6P63Q6KawE8T8E6h1EEDaEegcRhA2hBRgEYUMouARhQ+gzLkHYEPqMSxA2hN4qE4QNoeAShA2h4BKEDakouIqieJ555pn/GBsbm3zw4MEFv/nNb1aN98QIgvjnVLTk0ev1mj09PbmzZ89Of+6551ZYlqU/+uijHQAA8Xi8IZvNOqLRqPDWE8uyIJPJOKLRqPBVKs45XB4XdYUL6yYSCUilUihXkiRIp9Mot1AooM+Vruto1zTNqmqEdQHwNYrH4+gapdNptCvLMrq+mqYxkR12FQXX4/HwYDAIhw4dumbKlCmze3t7XZlMxllXV8cvXrw4Y2Rk5C+hUOiU6GQZY85YLLYyFAqdFHU5545YLPbZUCjULuoCAMRisZsxbiaTqY9EIrNDodB5UVeW5UAkErkuFAqdFXU1TfNFo9EloVCoU9Q1DMMzMjKyPBQKnRZ1LctyxWKxfwuFQh2ibpU1QrupVKohHA5PD4VCvaKuJEnBSCQyNxQKnRN1C4WCPxqNLgqFQt2ibrFYrDEMo6Zi4ZObnf/Zn6Zp7pdeemnj7t27V+fzef9bb721oPxvhmG49u/ff0Mlx/mvf6ZpOvbu3XsjxuWcw+7du//tarvJZDJ45MiRRRhXkiR/S0vLEoyrKErNwYMHl2HcYrHofvvtt1E1KpVKzn379q3Anuc9e/Zc9fqOjo7Wt7a2LsC46XS69qOPPlqMcfP5vPf999//DMZVVdXzzjvvXF/p7Sv6jOvz+cyvf/3r72zatKmtrq5Ou/feewfK/9bS0nJtMpkMFAoF4S71bW1t19XX16NWYbz++usLJUmqzWazlT9LXebNN9+8NpvNBlOplE/U9fv9hf7+/vpcLifcbDaRSEyLx+PX5PP5gKibTCad0Wi08ciRI9NE3Z6enoZUKjXz0KFDi0XdDz/88LpEIlE/NDQkfJ4PHz48P5FI+HO5XMW70IrFovuVV15ZPzg42JjNZmv27NmzoFLXMAznzp07by8UCt6enp5ZZ86cmVKpa5qm429/+9utmUxmUjgcnvz6668vqdRljDlef/311fF4fJosy/5du3YtNwyj4sf1W2+9tSoajc5ctmzZ0CuvvLKyEqeqq8qZTMb185///OZ33323/tixY7NF/bGxMfmNN96o+AR9ksmTJyc/+OCDxgMHDkwXdVVVzf32t79d1NLSMlPU/eCDD+Y88cQT/yuRSAh3az558uTCjz76KGAYhvDPSTz//PMr9+7dO3NoaEjYXbx4sdTV1WUODAwIdR8vFAqOp556as2Pf/zjFUNDQ0IdplVVdX7/+99f/cILL0x/9913Kw6fw+Fgb7zxRm1nZ+fy1tbWKTt27LjZsqyKQuBwOPiePXtqwuHwjIGBgZr33ntvRqXjut1ufvDgQWdfX9/syZMnD7z88surVFWt6AnH6XTyw4cPs+7u7jlutzv/+9//fm2xWKz4yaqtra3U2tq64Je//OXKZ5555tZUKvWp97eq4HLOHTU1NXzatGmqZVnCrdLnzZsnOxwOCzN2OBwOFAoF71133SX8GzLLly8vrlixYkSWZa+oe/jw4c9ec8019eFwuOIHRZkZM2ac8ng8+T/84Q83iLqKotTNnj1bb29v/6yo63a7rZ6enml33HHHBRHPsiyHx+Mxpk6dGjcMQ+gdRm1tLXv44YdPe73eWaVSqeIHsdfrZbNnz85xzt1z5szRZs2aVfGYHo+Hz5kzJ+fz+djq1avzIvMFALj22msll8vFu7u7F82dO/dibW1txVeL5s2bJzmdTmhqakpf7h9dcbbmzZuX6+zsrO3q6roxGAzOjsfjn/pOsKrgTpkyxbz77rv7a2pqGleuXBkV9fP5/PyGhobGoaEh4VevwcHB5bfddpsjn88Lv909dOhQ08yZM/3r1q0bE3V/+tOf7nn88cd3LVu2bFjU7e3tnTxp0qTGL3zhC0Oi7je/+c2OxsZG95o1a/pE3Uwm47r99tuH5s2bp4p49fX17Ctf+Upk1apVbPr06UnRcWfMmOFct27d8Pr16/srdXRdd8+ZM2fO0NCQV1GU4Pz58/sr/cWVUqnkmjlz5rUdHR3z+/v7rw0Gg/MZYxW9Wpum6ZwyZcq8s2fPLohGo9ffdtttHtM0K3IZY45gMHhdf3//tW+//fYta9eujXk8nlIlLgCA1+u9bs6cOXU7d+58+Rvf+MaeefPmFT/NcXBO64UJwm7QyimCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAh/xdGVFaQoPVOigAAAABJRU5ErkJggg==",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -431,14 +440,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"As the first move of the game, it's best to place the piece in the center of the board to maximize possible winning directions.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -459,8 +468,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -471,14 +482,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"As the game has just started, I will place my piece near Alice's to prevent her from forming a line and to start building my own line.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        8\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -499,8 +510,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -511,14 +524,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"My opponent placed his piece next to mine. I should block him and also try to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        6\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -539,8 +552,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -551,14 +566,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"I see Alice is trying to create a line horizontally. I need to block her next move otherwise she will have a chance to win. I will place my piece at [7, 5].\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        5\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -579,8 +594,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -591,14 +608,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [7, 9] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        9\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -619,8 +636,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -631,14 +650,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"Alice is trying to create a line horizontally again. I need to block her next move otherwise she will have a chance to win. I will place my piece at [7, 10].\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        10\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -659,8 +678,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -671,14 +692,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [6, 7] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        6,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -699,8 +720,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAoS0lEQVR4nO2deZRU5Z33v7eWruqu6o1ma/YAzdqgQkRahLyaIBgUJTYmDKhJJoszepwkJ2fmJCeZzMwx7zhhJmcySeBkToyJIoqihEUQorQL0DRNA71R0EtVL9VVXfu+33uf9w8t3hhZ6nmaJRd/n3P6eLDu5z5P3ae+tdz7PPcnMcZAEIS20N3oDhAEwQ8FlyA0CAWXIDQIBZcgNAgFlyA0CAWXIDQIBZcgNAh3cM+ePVvx4osv1g0PD5dciw4RBHFlJN4JGBs2bFhZWVkpT5gwIfmjH/2oKZPJGBhj+mvUP4L41GAwGGSDwaAUtC3vzmVZNsyZM8fr9XrLM5mMbu/evZ89ffr0M7W1tdW8+1JVFR0dHVi4cCGvCsYY2tracMstt3C7ANDa2irkxuNxeDwezJgxg9tNpVIYHBzErFmzuN1MJgO73Y65c+dyu7Is4/z585g/fz63q6oqOjs7sWDBAm6XMYb29vbrPr7RaBR+vx/Tp0/ndhOJBFwuF2pqarjddDqNvr4+zJkzh9vN5XKsuLj4/65fv357QQJjjOvv1VdfnfaNb3zjkebm5jGMMWQyGeO2bdvamAC5XI698MILIipTVZX9/ve/F3aff/55IdflcrEDBw4IuX6/n+3evVvIjUaj7LXXXhNyk8kk2759u5CbzWbZiy++KOQqisL+8Ic/CLkjGaPBwUF26NAhIdfj8bB9+/YJuaFQiL3xxhtCbjwel3fs2PEEKzCH3J+469ev71u/fn0fr0cQxNWDzioThAa5ocGVJOlGNk8QmuWGBpfRkkKCEIK+KhOEBqHgEoQGoeAShAah4BKEBqHgEoQGoeAShAah67gEoUG4g3vkyJHJP/rRjzZejcbpOi5BiME9V3nq1Klum82mqqoqqaqKs2fPThoYGDCdOXOGu3FVVTEwMAARlzEm7ALA4OCgkOv3+2G324XcSCQCh8Mh5CaTSfT19Qm52WwW/f39Qq6iKMLujRojj8cjfKxCoZDw+MbjceHxTafTyGQyBW/PHVxJkoyZTKYsEomYS0tL06qqSowxqKrKuyuoqnrhTwTRdv+8bZE2RdsdiauqqubckTzfvP9pOs48cAdXp9OZv/Wtb3UFAoGSysrKVG1t7aDNZsssWrSId1eQZRk2mw0iLvtoraeo29bWJuS63W4oiiLkBgIBRKNRITcWi2F4eFjITaVScDgcQm4ul8P58+eF3Pxa3us9Rk6nE3q9Xsj1er1Ip9NCbjgcRiAQEHITiQScTmfB23MHd8KECaG1a9c28HoEQVw96HIQQWgQCi5BaBAKLkFoEAouQWgQmjlFEBqE7oBBEBqEvioThAah4BKEBqHgEoQG4Q5uKBQyeb3eclmW6cwSQdwguKc8/va3v6222Wy3r1ix4sxXv/rV7mvRKYIgLg93cB977LHBp556an5ZWVkyX/Tr+PHjVr2ev2Cfqqo4fvw4jEYjt8sYQ1NTE0wmE7cLAE1NTTCbzdxeOBxGf38/wuEwtxuPx3H+/Hkkk0luN5VKoa2tDbIsc7vZbBYnT54UuvymKAoaGxthMHC/VC6Mb1FREbebH1+RMQoEAnC5XAgEAtxuNBpFb28vYrEYt5tMJtHZ2cm1PC9PJpPhGlvu0fjpT3+6pLi4eNq8efMaTSaTeu+997bG4/HEgw8+yLsrKIqCVCoFEZcxhng8LuQCH662EXGHh4fR3t6OlStXcrvBYBAVFRVYs2YNtxuLxWAymYT6nE6nAUDIlWUZmUxGyFVVFclk8rqP79DQEM6fP4977rmH2/X5fGhpacHq1au53UgkAovFgrVr13K7yWQSf/rTnwrenju4//qv/3omm82et1gsMQAwm805o9HIiouLeXcFWZZRVFQEEZcxNiLXaDQKuWazWbjdkbgjOVYAhJ9vLpcTbldV1U/VGGUymREdK52u8FNO3MGtrKxMAUjxegRBXD3ochBBaBAKLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQ7uAePXp09L59+xb7/X7+2d8EQVwVuKc82u32zHPPPTfrkUceSf/93/99pyzLelVVhVatKIoCRVGEXMaYsAt8ODf0evd5JK4syyNqV/T5jqRdVVVHNL4j6bMWx5fnHmzcwV26dGmqubnZm0wmjR8t61vc2NhYyrsf4P8v+xJZEgh8uDRPZEkgAOHlhJFIBAMDAwgGg9xuIpFAV1cX4vE4t5tOp9He3o5sNsvt5nI5tLS0CBejOn78ONcE+DyMMRw/flxoSSAgPr6hUAgulwter5fbjcVisNvtiEQi3G4qlcLZs2eRSvFP5ecdV+4junfv3s/o9frKuro6m8lkUtetW3dCluXYxo38JXNlWYZOp8Ojjz7K7ebfkUVdRVGEXLfbjdbWVqFlX4FAAEePHhVa9hWLxVBZWYn6+npuN5VKobi4GBs2bOB2c7kc9Ho9Nm3axO3m3yiu9xg5nU7YbDahpZderxfNzc1CSy/D4TAaGhqwbt06bjeRSODNN98seHvu4H7ve9/rBnBV7nxB91UmCDHovsoEoUHochBBaBAKLkFoEKodRBAahH7jEoQGoa/KBKFBKLgEoUEouAShQSi4BKFB6KwyQWgQ7uBGIhHdd77znQecTmfVSBuns8oEIQb3XOXdu3fPfPnll6d84xvfaK2urg729vaOc7vdxnPnznE3rqoq3G43RFzGmLALfFgDSMT1+/1wOp1CbiQSwdDQkJCbSCTgcrmE3EwmI3ysFEXR3Bh5PB7hMQoGg8JuLBYTHt9UKsW1Qog7uOFweNLMmTPHOByO6tmzZw+Gw2FLIpHQiyxzU1UV8XhcaIkc8OGLWdQVbTccDiMWiwm50WhU2E2lUsJ9zmazwq6iKMIuY2xEYyTqjmSMRuImEgnhY5VOp7mWXXIH9+mnnz58++23d86ZMyduNBrZ4sWL7Xa7PX3nnXfy7gqyLMPhcEDEZYyhu7tb2O3q6hJy3W43DAaDkBsIBJDNZoXcWCyGUCgk5KZSKbjdbiE3l8uhv79fyFVVFb29vdd9jJxOJ0pKSoRcr9cLxpiQGw6HEY/HhdxEIgG/31/w9kIrnOvq6jwiHkEQVwe6HEQQGoQuBxGEBqFFBgShQeirMkFoEAouQWgQCi5BaBAKLkFoEDqrTBAahM4qE4QG4Q5uf39/9YEDB5bG4/Hia9EhgiCuDPeUx/fff39+c3Ozua6ujgHAR0W/pFwux914vkiSiJsvUSHi5tsWcfNFsMi9MvmiX6JjpKqqpp7vSN1rWvRr0qRJp48cObLoueeem/fUU0+d+ajol1Xk92q+oJRIUSjGGJqamlBUVMTtAsCJEydgMpm4vXA4jIGBAYRCIW43Ho+jq6sLiUSC202n02hraxN6UeRyOZw8eZLbAz58gxMtzJYfI9HCbKLjGwwG4XK5uCbt54lGo7Db7YhGo9xuMpnE2bNnkU6nud1MJsO1PXdizpw5U2GxWMrr6uo6r0bRL0mShApK5Yt+ibqyLAu5N7LoV0VFhXDRL7PZfEOKfjHGrvsYUdGvi/Dd7363F0Avr0cQxNWDruMShAah4BKEBqHgEoQGoeAShAah4N7E5M/MptNp5HI55HI5mq12kyB0zynirxvGGCKRCE6ePAmPxwNJkpBKpfDqq6+irKwMixcvRnV1Nc0V1zAU3JsMxhjOnDmDjo4OLFmyBCtWrLgwwUVVVQwPD+Po0aOoqqr62GOEtqCvyjcR+dAODAzgkUcewaxZs1BUVASdTgedTgeDwYBJkybhS1/6EkwmE95++22ue/kSfz1QcG8iQqEQOjo6cN9998FkMl3yq7Ber0ddXR1kWUZPTw/97tUgFNybiJMnT6Kurq6gucE6nQ4rVqxAS0sLfepqEO7gvvvuu2P+6Z/+ae2uXbvGXosOEWIoigKPx4OpU6cWfNLJarVCp9MhlUpd494RVxvuMxP/8z//s1SSpOLbbrtNUhRF6unpGe9yuYpsNht344qiwOVyQcRljAm7wIeLBURcn8+HgYEBITccDmNwcFDITSQSFybPX4xUKgVFUbhONkmSBIvFgpaWFowde/H3YVmWhY+zqqojGl/RMRoeHhYeo2AwKDxGsVjssmN0Oa550a8pU6ZkAKQ7Ojrm19fXe2OxWHEqldJFIhHeXUFVVSSTSYi4jDGkUikhF4CwG4vFhPs8EjeZTF62z5lMBrIsc+9XURQkEolL7ldRFBqjAonH48Lu9Sj61Xjy5Mn5NTU1NoPBwG677TZHT09PeunSpby7gizL6O3thYibLwol6p47d07Idbvd0Ov1Qm4gEEA6nRZy8xXkLuWqqorBwUFks9mC1xnnq+k99NBDKCkpueg2uVwODodDqM+qqqK7u/u6j5HT6URxcbGQ6/V6oSiKkJuv9CfiJhIJeL3egrfnDu706dNj06dPP87rEdcWSZIwZcoUdHV1oba2tqDfucFgEEajEWaz+Tr0kLia0FnlmwRJkrBo0SKcOnUKyWTyitvLsoyGhgYsXboUOh29DLQGjdhNhMViwZ133ok9e/YgGo1e9PosYwyZTAaHDh3ChAkTMGnSpBvQU2Kk0Hy3mwhJkjBz5kwYDAbs2rULNTU1mDt3LsrKyiBJEhKJBBwOB86cOYPa2lrceuutNF9Zo1BwbzIkScK0adNQXV2Nc+fO4fDhw4hGoxgaGkJNTQ0mT56MdevWwWq1Umg1zA0N7qfhhcMYQy6Xg8vlwvnz53Hu3DnMnj0bkydPRnFx8WWPQd71er3o7+9Hf38/7HY7qqurYTabL+lKkgSTyYR58+Zh9OjRGBwchMfjwe23335F9y/77HQ60d3djfPnz3P32ePxYHBwsKA+/2W7LpcLPT09OH/+PKZMmXJF99PGDQ3uzTxHNn/f53fffRdbtmxBU1MTfD4fFEXB5s2bMX/+fHz9619HfX39J8KQd0+ePImuri6MHTsW48aNw+c//3n09fXh2LFjGDNmDJYvX35RV5ZltLS0fMxdvXr1BXfs2LG46667LhrCXC53oc8nTpy40OctW7agtrYWf/u3f4svfelLBbe7Zs2aK7abdxsaGrB169aPtfub3/zmsu1+amGMjegvk8kYt23b1sYEyOVy7IUXXhBRmaqq7Pe//72w+/zzzwu5LpeLHThw4IrbJZNJ9s///M+svLycAbjon8lkYps2bWLDw8OfcF999VV25MgRlk6nmaqqH+t7LpdjnZ2d7A9/+APz+Xwfe7wQt6Oj45LuT37yE1ZWVnbZPm/cuJF5PJ5PuDt27CioXb/ff+FxVVVZIpFgP/7xj6/Y7qOPPsq8Xu/H9n0xBgcH2aFDh644RhfD4/Gwffv2CbmhUIi98cYbQm48Hpd37NjxBCswd/Qb9xqQ/6R49tlnLzuNLZPJYPv27TAYDNiyZQuKi4uhKAoOHTqEefPmYd68eZ/4dJEkCQaDAXPnzsWoUaNw4MABrF+/HmazGYqi4ODBg1d0582bh6qqqk+4W7duxbPPPnvZm3NnMhm8/PLLKCoqwpYtW2A2m6GqKt566y3U1tZi7ty5l2131KhR2L9//4V2GWPYunUrfvazn12x3Zdeegkmkwm//OUvP/XXnuly0FWGMYbz589j8+bNBc09VVUVO3bswP79+8EYg8PhuPAiv9xXQkmSMG7cOCxYsADHjx8HYwx2ux1FRUVcblNT08f6XMgd9VVVxSuvvIK33noLjDH09PTAbDZjzpw5V2x3/PjxmD9/Pk6cOAHGGGw2G1e727Ztw6FDh27qn1mFwB3cWCxmDIVCpalUSqyuxKeAXbt2we12F7x9KpXCCy+8AEVR0NraiiVLlhT0O06SJMybNw8DAwOQZVnYzeVy2LlzJzweD1efX3zxRWSzWbS2tuKOO+4oaCKHJEmora1Ff38/stksXn/9da6pful0Gtu2bROal30zwR3ctra2cU888cSXX3rppdpr0SGto6oqjh49yv2JcPbsWXi9XiSTSVRVVRXsGY1GlJWVwe/3I51OY9SoUVyu1WpFIBDAyZMnuft8+vRpBINB5HI5lJeXc7VbWlqKUCiEI0eOcLd76tQp4YULNwvcv3GXLVvm/NnPfhZevnx5byaT0R04cOC2EydOWEUKaKmqiubmZhQX81fsZIyhubkZFouF2wU+XHRutVq5vXA4jL6+PsTj8Ys+nsvl0NfXx71fj8eDV155BSaTiWsKoiRJMBqNeO2112A0GrnOuEqSBL1ej9dee014meLOnTsve7eNy7X9xhtvoKuri7tdj8eDN95445JvUoFAAENDQ0Lhjkaj6OnpEVqjnEwm0dHRAUVRuN1MJsNV+Is7uF1dXVaz2SzPnj07CgArVqywBQKB5KpVq3h3BUVREIlEIOIyxhAMBoVcAPD7/UKux+NBe3s7vvCFL1z0cUVR8Nxzz3EHoaqqCmvWrEFjYyNUVS04vOyja5+rV6/G8eN8az/YR5ed7rvvPuzfvx89PT1cfmlpKVatWnXh9ypv2/feey927NiB/v5+LreqqgorV67E6NGjL/r40NAQurq6cPfdd3PtF/jwdVFWVib02ohEIjAYDEJuMplEQ0NDwdtzB3fixInJX/7yl2/l/221WtNms1ktLS3l3RVkWYbZbIaIyxi7IW48HkdxcfElXcYY6urq8M4773Dtd/bs2Zg+fTra2toQDocL/soryzKi0SimTZuGM2fOIBKJoKKiomA3Ho9jypQpWLx4MQ4ePMjV5wULFmDy5Mk4ffo04vF4wV+XFUVBPB7HxIkTcdddd+GDDz7gCv6tt96KiRMnXrIEp9VqvewYXY5UKiXsKooi7OZv6Ffw9rwNWCwWdezYsfwFQC/CzXoh/eGHH8aYMWMK3r6oqAibNm2CwWBAbW0tWlpaCn4h9/T0oLq6GkajEbW1tWhubi7IZYyhu7sb1dXVKCoqQn19Pddv66KiImzcuBFms5m73XPnzl2YSfXwww+jsrKSq90NGzYI19y9Wbihl4NuxlP6+bOmTz31VEG3kZEkCQ888AAefPBBSJKEmpoahMNh2O32yx6f/E+FEydO4M4774QkSZg1a1bBbigUQnNzM+rq6iBJEhYsWIAnn3yyoEBIkoS1a9figQcegCRJmD17NoLBIBwOR0HttrS0YOnSpZAkCQsXLuRq96GHHsKaNWtu2jf9Qrmhwb1ZD77BYMB3v/tdPPHEE5c98abX6/HFL34RP//5zy98vTIYDLjvvvtw/PhxnDlzBrIsfywM7KOC3v39/dizZw9Wrlx54e4Vf+kqinJRt6+v76Lu97//ffzd3/3dZSc35Nv4+c9/fuHEYP7/HTt2DK2trZds1+FwYM+ePbj33nsvtKvX6/H9738f3/72t6/Y7v3334//+q//Ej4heTNBc5WvEVarFZs3b8by5cuxdetWtLa2IhKJgDEGq9WK6dOn4/HHH8fXvva1T/w2tFgsqK+vx9GjR/HKK69gypQpqK6uhl6vh8/nQ19fH0wmE9atW3dhyd7F3JdffpnLtVqt+I//+A8sW7YMv/71r9He3v6xPs+YMQNf/epX8fjjj6O8vPwT7uXadTgcKC4u/kS7kiShtLQUmzdvxrJly7B161audj+t0JTHa4QkSTCbzVi/fj3WrFkDu92Ojo4OdHZ2Yu3ataipqUF5eflFT0jkV/fcfffdSCaTcDqdcDgcaGlpwf33349Vq1ahrKysYHdgYACNjY148MEHr+jm+/zFL34Rdrsddrsd77//Pr7yla9csc9msxn33HPPhXZdLhcaGhrw8MMPY/Xq1Vds98tf/jLuv/9+2O129PX14Z133sHGjRsv2+6nFQruNSZ/C9QFCxZgwoQJsFgsWLJkCZc7e/ZsTJgwAdFoFAsWLOB2p0yZAq/Xy+VarVYsXLgQc+fORTweF+pzTU0N+vv7hdpdsGABgsFgwe1+2qC3MILQIBRcgtAgFFyC0CAUXILQINzBHRgYKHv77bdv6erquvit7zmg0/oEIQb3WeWf/OQnt7/33nuznnjiCd0//uM/nlZVVcpfYOdFVVWIuvlbeIiWiLwR7Y7EvVHHaiTtarHPIx1fVVVH1OdC4Q6uoiisqqoqYDAYTJlMRrdr164lR48eLRXtbGNjI7cHfHiQjh07Jnxt79ixY9Dr9dxeOBzGwMAAfD4ftxuPx9Hd3S1cFKq9vV1ouVk2m8WpU6eEi4Ll75LBi6qqOHbsmNA3K8YYGhsbhcYoGAzC7XZjeHiY241Go7Db7QgGg9xuMpnE2bNnkUgkuN1MJsP1WuYO7vLly4PxeHx8ZWWl12QyqQ8//HCTqqqxjRs38u4KsixDr9fj0Ucf5XbzL6THHntMyFVVVch1u91obW3F6tWrud1AIICjR49i7dq13G4sFsOoUaNQX1/P7aZSKVgsFmzYsIHbzeVyMBqN2LRpE7ebX54oOkaMMSE3X+py5cqV3K7X60VzczPWrFnD7YbDYTQ0NGDdunXcbiKRwJtvvlnw9tzB/eY3v3nmm9/85hlejyCIqwedVSYIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1Cd3kkCA3CHdy9e/fOePrpp9e2tbUVfhfrS3Az3+WRIK4l3HOVt2/fXuN0Otmrr766aOHChe8MDAyM9vl8Rrvdzt24oijwer0QcRlj8Pl8Qi4AYdfn82F4eFjIDYfD8Hg8Qm4ikRA+VplMRtiVZVlzYzQ8PCw8RsFgUNiNxWLC45tKpZDL5Qrenju4Tz/99MmtW7feYzQafblcTnK5XJWBQMAwODjIuyuoqopgMAgRN38nfxEXgLAbDAbh8/mE3Gg0Kuwmk0n4/X4hN5vNIhAICLmKooxojETbBcTHyOfzCR/ncDgsfJzj8biwm06nuZZdcgd39OjR+rvvvtu1cuXKM0ajkS1durR7cHAw9bnPfY53V5BlGU6nEyIuYwz9/f3CrsPhEHLdbjeKi4uF3EAgAMaYkBuLxZBIJITcVCoFv98v5OZyObhcLiFXVVUMDAxc9zFyOp0oLy8Xcr1eLwwGg5AbDoeRzWaF3Gu+rK+mpsZTU1NTeOlygiCuOnQ5iCA0CAWXIDQIXcclCA1C9XEJQoPQJy5BaBD6xCUIDUInpwhCg1BwCUKD0G9cgtAgBQU3k8noGxoabvF4PJa33357TlNT07ir0Tj9xiUIMQqa8qgoCv793/995pNPPmnZsWPH2EwmU/Taa6+9ptPpmKqqOsYYFEXhbjxfIEnEBT4M/vV2R9Lnkboj6fONeL75Ui83Yoy0Or6FUlBwS0pKlBkzZrgBmJcsWeLv7++fCAAfFf26faRFv0SLQo2k6JdoQalIJIL+/n74/X5uN1/0KxqNcrsjKfqVy+XQ0tIyoqJfItyoMRpp0S+Hw4FQKMTtJpNJ2Gy2v56iX+l02mixWGZ0dHREHA6Hefz48S6dTsdMJhOjol+FcyOLfpWUlNyQol+SJF33MaKiXx9hNptz//mf//kid28Igrgm0FllgtAgNHOKIDQITcAgCA1CwSUIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1CwSUIDVJQcGOxmPEHP/jB+tbW1gn/8A//8PCRI0fGXOuOEQRxaQqaq1xaWpobHBz0MsbM1dXV/nPnzlnvuusuHwD09/eP8Xq9xt7eXu7G80W/RFzGmLALfFhfRsT1+Xxwu91CbjgcxvDwsJCbSCTg8XiE3HzRLxE3X/RLxFVV9YaM0fDwsPAY5VcWibixWEx4fK9Z0S+TyWRMp9PmiooKYzabNQNALpeThoeHy0OhkGFoaIi7s6qqIhQKQcRljAm7AITdYDCIQCAg5EajUWE3mUwiGAwKudlsVtjNF/3S0hj5fD74/X4hNxwOC49RPB4XPlbXpOhXJpMxrFy5Uh0aGrJWVFSoo0aNsgCA0Whkd9xxR8/AwEBqxYoV3J2VZRmDg4MQcRlj6OvrE3btdruQ63a7YTabhdxAIABVVYXcWCyGeDwu5KZSKfh8PiE3l8thaGhIyFVVFf39/dd9jJxOJ8rKyoRcr9cLvV4v5IbDYWQyGSH3mizrM5lM8le+8pXD3L0hCOKaQGeVCUKDUHAJQoPQQnqC0CC0kJ4gNAh9VSYIDULBJQgNQsElCA1CwSUIDULBJQgNUnBwE4mEWZZlXTweNyeTSf66EARBXDUKCm4kEimqr6/fdOzYsZof/OAH/+fpp5++91p3jCCIS1PQXOXy8vLslClTbKNHj87W19fbfvWrX80DgGw2qzt06NDCU6dOWaxWK3fjqqri1KlTKCsr43YZYzh16hQqKiq4XQA4ffo0Kisrub1QKASHw4FMJsPtxmIx2Gw2oevXqVQKp0+fhtFo5Haz2SxaWlpQUlLC7SqKgpaWFpSWlnK7+TEqLy/ndgHxMQoEAnA6nUgmk9xuJBJBd3e3UIG0RCKB9vZ2oSJnmUyGq1hYQcHNZrP6eDw+5d133y393e9+t/Cpp556EwCKiorUJUuW9LpcrtTy5cu5O6soCjweD0RcxhhcLpeQC3y4gkTE9Xq9sFgsQm4oFIJOpxNy4/E40um0kJvJZBCJRIRcWZbh8/mEXFVV4Xa7hcdoaGhIyHW5XKisrBRy/X4/ioqKhNxoNApFUYTcZDKJDz74oODtCwquXq9Xf/rTn76nKApbtWpVp9lsvvDWUFFRkbBYLOqoUaO4OyvLMiwWC0RcxtgNcTOZDEpLS4XbtVqtQq7RaBR2U6mU8PPN5XLCrqqqIxqjkpISITeZTAofK1mWhcdXp9MJt2symbhKihYaXDZt2jQXd28Igrgm0OUggtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIAUFN5vN6pqamub4/X5rU1PT7NbW1tFXo3G6yyNBiFHQlMdcLif98Ic/XPgv//Ivyrlz5yr27dt32+7du18BAMaY9NF/uRtnjF34E3FF2/3LfYh417tdLT7fkYzvn+9D1LkR4zvS51soBQXXYrEos2bNclZWVso1NTUBSZKmAUAmk9G9/vrrd3zwwQelIsugVFVFY2Oj8OAcPXqU2/tzV+QTPxwOY3BwEF6vl9uNx+Po7u5GKBTidtPpNNrb27mWfuXJZrM4deoUstkst6soCpqamqCqKrebH18RRjJG+Yp7Lhf/9PpoNAqHw4FAIMDtJpNJ2Gw2xGIxbjeTyXAtBywouOl02mgwGGr27NkzVqfTKTNnzgwAgMlkUuvr648zxmIbN27k7qwsyzAYDHj00Ue5XcYYJEnC448/LuQCEHLdbjdaW1uxevVqbjcQCODo0aNYu3YttxuLxXDw4EHU19dzu6lUClarFRs2bOB2c7kcioqKsGnTJm5XVVXo9Xo89thj3O5IxsjpdMJms2HlypXcrtfrRXNzM9asWcPthsNhNDQ0YN26ddzuNSn6ZTabc7/4xS9eyB9MvV5PdzIniBtIwfVxdTodhZUg/kqg2kEEoUGodhBBaBCagEEQGoSCSxAahH7jEoQGod+4BKFB6KsyQWgQCi5BaBAKLkFokIKCG4/Hjc8888xDHo9n1KFDh6b/+te/XnytO0YQxKUpaMqjyWSSOzs7Ix0dHWN/8YtfLFQUJfPkk0+2AIDb7a4IhUKS0+nkXnqiKAqCwaDkdDq5z1IxxvBRu0JnuERdr9cLv98v5IbDYQQCASE3kUgIH6tMJiPsyrI8ojESdQHxMXK73cJjFAgEhN1YLCY8vqlUSuVZYVdQcI1GI7NarTh8+PDEqqqqCV1dXfpgMKgrLS1lPT0944aGhl6y2WyneDurqqrO5XItstlsJ3ldxpjkcrk+a7PZmnldAHC5XLeLuMFgsGxwcHCCzWY7x+vGYjHL4ODgZ2w2Wwevm0qlzE6nc7bNZmvldbPZrHFoaKjWZrOd5nUVRdG7XK5bbTZbC687wjESdv1+f8XAwMBYm83WxeuGw2Hr4ODgFJvNdpbXTSQSxU6nc6bNZmvnddPpdFE2my0qWPjzxc6X+kulUobf/e53q3fv3n1HNBot/uMf/zg9/1g2m9W/+eabCwrZz1/+ybIs7d279xYRlzGG3bt333q9XZ/PZz1y5MhMETccDhc3NDTMFnHj8XjRoUOH5om46XTasH//fqExyuVyun379i0UPc579uy57uM7PDxc1tjYOF3EDQQCJe+///4sETcajZrefvvtuSJuMpk0vvXWW/ML3b6g37hms1n+2te+9tbatWubSktLUw8++KA9/1hDQ8Nkn89nSSQS3FXqm5qaPlNWViY0C2Pnzp0zwuFwSSgUKvxd6iN27do1ORQKWf1+v5nXLS4uTvT29pZFIhHuYrNer3eM2+2eGI1GLbyuz+fTOZ3OyiNHjozhdTs7Oyv8fv/4w4cPz+J133333c94vd6yvr4+7uP83nvvTfN6vcWRSKTgVWjpdNqwbdu2FQ6HozIUChXt2bNneqFuNpvVbd++/a5EImHq7OysPnPmTFWhrizL0iuvvHJnMBgsHxgYGLVz587Zhbqqqko7d+68w+12j4nFYsU7duyozWazBb+u//jHPy52Op3j582b17dt27ZFhTgjOqscDAb1mzdvvv3gwYNlx44dm8Drezye2Ouvv17wAfpzRo0a5XvnnXcqDxw4MJbXTSaTkS1btsxsaGgYz+u+8847k773ve993uv1cldrPnny5Iz333/fks1muW8n8atf/WrR3r17x/f19XG7s2bNCre1tcl2u52r+ngikZB++MMfLv23f/u3hX19fVwVppPJpO7HP/7xHb/97W/HHjx4sODwSZKkvv766yWtra21jY2NVS+++OLtiqIUFAJJktiePXuKBgYGxtnt9qI//elP4wpt12AwsEOHDum6u7snjBo1yv78888vTiaTBb3h6HQ69t5776nt7e2TDAZD9De/+U1dOp0u+M2qqakp19jYOP2///u/Fz3zzDN3+v3+Kz7fEQWXMSYVFRWxMWPGJBVF4S6VPnXq1JgkSYpI2wMDA5ZEImH6whe+wH0Pmdra2vTChQuHYrGYidd97733Pjtx4sSygYGBgl8UecaNG3fKaDRG//d//3cBrxuPx0snTJiQaW5u/iyvazAYlM7OzjGf+9znzvN4iqJIRqMxO3r0aHc2m+X6hlFSUqJ+61vfOm0ymapzuVzBL2KTyaROmDAhwhgzTJo0KVVdXV1wm0ajkU2aNCliNpvVO+64I8rTXwCYPHlyWK/Xs/b29plTpkzpKSkpKfhs0dSpU8M6nQ41NTWBj+pHF5ytqVOnRlpbW0va2tpusVqtE9xu9xW/CY4ouFVVVfKqVat6i4qKKhctWuTk9aPR6LSKiorKvr4+7k8vh8NRu2zZMikajXJ/3T18+HDN+PHji5cvX+7hdZ999tk93/nOd3bMmzevn9ft6uoaVV5eXnnffff18brf/va3WyorKw1Lly7t5nWDwaD+rrvu6ps6dWqSxysrK1P/5m/+ZnDx4sXq2LFjfbztjhs3Trd8+fL+FStW9BbqZDIZw6RJkyb19fWZ4vG4ddq0ab2F3nEll8vpx48fP7mlpWVab2/vZKvVOk1V1YI+rWVZ1lVVVU3t6OiY7nQ65y9btswoy3JBrqqqktVq/Uxvb+/k/fv3L6mrq3MZjcZcIS4AmEymz0yaNKl0+/btz3/961/fM3Xq1PSVHIkxmi9MEFqDZk4RhAah4BKEBqHgEoQGoeAShAah4BKEBqHgEoQGoeAShAah4BKEBqHgEoQG+X/Dcc4OaIWRHwAAAABJRU5ErkJggg==",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -711,14 +734,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"Alice is trying to build a vertical line now. I need to block her next move otherwise she will have a chance to win. I will place my piece at [5, 7].\",\n",
       "    \"move\": [\n",
       "        5,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -739,8 +762,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -751,14 +776,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [8, 7] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        8,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -779,8 +804,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -791,7 +818,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"Alice is trying to build a vertical line again. I need to block her next move otherwise she will have a chance to win. I will place my piece at [9, 7].\",\n",
       "    \"move\": [\n",
       "        9,\n",
@@ -824,17 +851,13 @@
     "            msg = player(msg)\n",
     "            \n",
     "            i += 1"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:10:33.989202Z",
-     "start_time": "2024-03-27T08:09:59.446824Z"
-    }
-   }
+   ]
   },
   {
    "cell_type": "markdown",
+   "metadata": {
+    "collapsed": false
+   },
    "source": [
     "## Future Direction\n",
     "\n",
@@ -845,10 +868,7 @@
     "- Fine tune a model to improve the performance of the agent\n",
     "\n",
     "For complete code, we provide in [code/game_gomoku.py](./code/game_gomoku.py)."
-   ],
-   "metadata": {
-    "collapsed": false
-   }
+   ]
   }
  ],
  "metadata": {

From fbca61ffdc590a7cd20492684114f4574febd142 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 22 May 2024 20:42:37 +0800
Subject: [PATCH 27/55] revert back unnecessary changes

---
 examples/game_gomoku/main.ipynb | 260 +++++++++++++++-----------------
 1 file changed, 120 insertions(+), 140 deletions(-)

diff --git a/examples/game_gomoku/main.ipynb b/examples/game_gomoku/main.ipynb
index 569321c15..7044a8243 100644
--- a/examples/game_gomoku/main.ipynb
+++ b/examples/game_gomoku/main.ipynb
@@ -31,13 +31,6 @@
   {
    "cell_type": "code",
    "execution_count": 1,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:57.602994Z",
-     "start_time": "2024-03-27T08:09:57.565787Z"
-    },
-    "collapsed": false
-   },
    "outputs": [],
    "source": [
     "YOUR_MODEL_CONFIGURATION_NAME = \"{YOUR_MODEL_CONFIGURATION_NAME}\"\n",
@@ -46,38 +39,38 @@
     "    \n",
     "    # ...\n",
     "}"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:57.602994Z",
+     "start_time": "2024-03-27T08:09:57.565787Z"
+    }
+   }
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "## Step 1: Prepare a board for Gomoku\n",
     "\n",
     "First we create a `BoardAgent` class by inheriting from `AgentBase`, which manage the game board and update the game status as follows.\n",
     "\n",
     "To create a better visual experience, we also provide a function `board2img` to convert the board to an image. "
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   },
   {
    "cell_type": "code",
    "execution_count": 2,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:58.153932Z",
-     "start_time": "2024-03-27T08:09:57.571266Z"
-    },
-    "collapsed": false
-   },
    "outputs": [],
    "source": [
     "import numpy as np\n",
     "import matplotlib.pyplot as plt\n",
     "import matplotlib.patches as patches\n",
     "\n",
-    "def board2img(board: np.ndarray, save_path: str)->str:\n",
+    "def board2img(board: np.ndarray, output_dir: str)->str:\n",
     "    size = board.shape[0]\n",
     "    fig, ax = plt.subplots(figsize=(10, 10))\n",
     "    ax.set_xlim(0, size - 1)\n",
@@ -107,30 +100,29 @@
     "    ax.set_xticklabels(range(size))\n",
     "    ax.set_yticklabels(range(size))\n",
     "    ax.invert_yaxis()\n",
-    "    plt.savefig(save_path, bbox_inches='tight', pad_inches=0.1)\n",
+    "    plt.savefig(output_dir, bbox_inches='tight', pad_inches=0.1)\n",
     "    plt.close(fig)  # Close the figure to free memory\n",
-    "    return save_path"
-   ]
+    "    return output_dir"
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:58.153932Z",
+     "start_time": "2024-03-27T08:09:57.571266Z"
+    }
+   }
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "The following code shows the implementation of the `BoardAgent` class. The agent manages the game board and updates the game status."
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.392069Z",
-     "start_time": "2024-03-27T08:09:58.159470Z"
-    },
-    "collapsed": false
-   },
    "outputs": [],
    "source": [
     "import numpy as np\n",
@@ -237,13 +229,18 @@
     "            else:\n",
     "                count = 0\n",
     "        return False\n"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.392069Z",
+     "start_time": "2024-03-27T08:09:58.159470Z"
+    }
+   },
+   "execution_count": 3
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "### Step2: Prepare a Gomoku Player Agent\n",
     "\n",
@@ -253,18 +250,13 @@
     "2. Within the agent, to enable the agent to think, we ask LLMs to respond in a dictionary format, which contains the thought and the move (\"thought\" field must come before \"move\"). To achieve this, we prepare a parsing function to extract the dictionary from response. \n",
     "\n",
     "The implementation of the Gomoku agent is as follows: "
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.397713Z",
-     "start_time": "2024-03-27T08:09:59.392729Z"
-    },
-    "collapsed": false
-   },
    "outputs": [],
    "source": [
     "import json\n",
@@ -335,7 +327,15 @@
     "        \n",
     "        # Hide thought from the response\n",
     "        return Msg(self.name, response[\"move\"], role=\"assistant\") \n"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.397713Z",
+     "start_time": "2024-03-27T08:09:59.392729Z"
+    }
+   },
+   "execution_count": 4
   },
   {
    "cell_type": "markdown",
@@ -346,25 +346,17 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.446367Z",
-     "start_time": "2024-03-27T08:09:59.398116Z"
-    },
-    "collapsed": false
-   },
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m2024-03-27 16:09:59.427\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models\u001b[0m:\u001b[36mread_model_configs\u001b[0m:\u001b[36m171\u001b[0m - \u001b[1mLoad configs for model wrapper: post_api\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.435\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m_create_monitor_table\u001b[0m:\u001b[36m341\u001b[0m - \u001b[1mInit [monitor_metrics] as the monitor table\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.435\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m_create_monitor_table\u001b[0m:\u001b[36m342\u001b[0m - \u001b[1mInit [monitor_metrics_quota_exceeded] as the monitor trigger\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.436\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m311\u001b[0m - \u001b[1mSqliteMonitor initialization completed at [./runs/run_20240327-160958_kbf2ta/agentscope.db]\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.440\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models.model\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m257\u001b[0m - \u001b[1mInitialize model [post_api]\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.441\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models.model\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m257\u001b[0m - \u001b[1mInitialize model [post_api]\u001b[0m\n"
+      "\u001B[32m2024-03-27 16:09:59.427\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models\u001B[0m:\u001B[36mread_model_configs\u001B[0m:\u001B[36m171\u001B[0m - \u001B[1mLoad configs for model wrapper: post_api\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.435\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m_create_monitor_table\u001B[0m:\u001B[36m341\u001B[0m - \u001B[1mInit [monitor_metrics] as the monitor table\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.435\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m_create_monitor_table\u001B[0m:\u001B[36m342\u001B[0m - \u001B[1mInit [monitor_metrics_quota_exceeded] as the monitor trigger\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.436\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m311\u001B[0m - \u001B[1mSqliteMonitor initialization completed at [./runs/run_20240327-160958_kbf2ta/agentscope.db]\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.440\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models.model\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m257\u001B[0m - \u001B[1mInitialize model [post_api]\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.441\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models.model\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m257\u001B[0m - \u001B[1mInitialize model [post_api]\u001B[0m\n"
      ]
     }
    ],
@@ -389,13 +381,18 @@
     ")\n",
     "\n",
     "board = BoardAgent(name=\"Host\")"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.446367Z",
+     "start_time": "2024-03-27T08:09:59.398116Z"
+    }
+   },
+   "execution_count": 5
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "### Step 3: Start the Gomoku game\n",
     "\n",
@@ -404,32 +401,26 @@
     "In this game, we use a message hub to share messages between the two players (board agent doesn't have memory), so that one player can hear what the board agent says to the other player, and what moves the other player makes. \n",
     "\n",
     "Note here we only show 10 steps of the game. You can adjust the `MAX_STEPS` to play more rounds."
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   },
   {
    "cell_type": "code",
    "execution_count": 6,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:10:33.989202Z",
-     "start_time": "2024-03-27T08:09:59.446824Z"
-    },
-    "collapsed": false
-   },
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: Welcome to the Gomoku game! Black player goes first. Please make your move.\n"
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: Welcome to the Gomoku game! Black player goes first. Please make your move.\n"
      ]
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -440,14 +431,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"As the first move of the game, it's best to place the piece in the center of the board to maximize possible winning directions.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -468,10 +459,8 @@
     },
     {
      "data": {
-      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAgxElEQVR4nO2de3Ab1d33f7pZkiU7tnO1CUkIcZKaJDwkJSSQhJe+hECZN7yZOm1TAr0MUKYwKeWPdmDo9DLwttP2j9LSdtrh0hYaGgiFXCAkQEwgjmMcJ7EdR4kdy7IkS7auK+1qV3s75/0jiIfnmYeic2THXfr7zHgYJvrsOdqfvrrs7tmfjVIKCIJYC/tUTwBBEHYwuAhiQTC4CGJBMLgIYkEwuAhiQTC4CGJBMLgIYkGYg3v27Nm6559/fu3Y2Fj1ZEwIQZBPx8Z6Aca2bds21tfXG01NTfJjjz3Wqaqqk1LqmKT5Ici/DU6n03A6nWZZj2XduGEYzqVLlyYSicQ0VVXt+/bt+/ypU6ceX7ZsWSPrtgghcObMGVixYgWrCpRS6O3thauvvprZBQDo6enhciVJgvHxcbjyyiuZXUVRIBKJwOLFi5ldVVUhGAzC5z73OWbXMAw4f/48XHXVVcwuIQT6+/th+fLlzC6lFPr6+i55ffP5PKRSKVi4cCGzWygUIBaLQXNzM7NbLBYhFArB0qVLmV1d16nX6/1/W7du3VmWQCll+nvppZcW3HPPPV/u6uqaSSkFVVVdL7zwQi/lQNd1+te//pVHpYQQ+uc//5nbfe6557jcWCxGDxw4wOWmUim6Z88eLjefz9OXX36Zy5Vlme7cuZPL1TSNPv/881yuaZr0L3/5C5dbSY0ikQg9dOgQlzs+Pk7379/P5WazWfqPf/yDy5Ukydi1a9f9tMwcMn/ibt26NbR169YQq4cgyMSBR5URxIJMaXBtNttUDo8glmVKg0txSSGCcIFflRHEgmBwEcSCYHARxIJgcBHEgmBwEcSCYHARxILgeVwEsSDMwT169Ojljz322J0TMTiex0UQPpivVZ4/f348EAgQQoiNEAJnz56dGw6H3adPn2YenBAC4XAYeFxKKbcLABCJRLjcVCoFwWCQy83lcjA8PMzlyrIMoVCIy9U0DUZGRrhc0zS53amq0fj4OPe+ymaz3PWVJIm7vsViEVRVLfvxzMG12WwuVVVrc7mcp6ampkgIsVFKgRDCuikghHz0xwPvuB8fm2dM3nErcQkhlnMreb4l/99pP7PAHFy73e657777BtLpdHV9fb2ybNmySCAQUFeuXMm6KTAMAwKBAPC49MO1nrxub28vlxuPx8E0TS43nU5DPp/nckVRhLGxMS5XURQYHh7mcnVdh/Pnz3O5pbW8l7pG0WgUHA4Hl5tIJKBYLHK5giBAOp3mcguFAkSj0bIfzxzcpqam7ObNm9tYPQRBJg48HYQgFgSDiyAWBIOLIBYEg4sgFgSvnEIQC4J3wEAQC4JflRHEgmBwEcSCYHARxIIwBzebzboTicQ0wzDwyBKCTBHMlzw+/fTTjYFA4NoNGzac/sY3vjE4GZNCEOSfwxzcu+++O/Lggw9eVVtbK5eafh0/ftzvcLA37COEwPHjx8HlcjG7lFLo7OwEt9vN7AIAdHZ2gsfjYfYEQYCRkREQBIHZlSQJzp8/D7IsM7uKokBvby8YhsHsapoGJ06c4Dr9ZpomdHR0gNPJ/FL5qL5VVVXMbqm+PDVKp9MQi8UgnU4zu/l8HoaGhkAURWZXlmXo7+9nWp5XQlVVptoyV+OJJ55Y7fV6F7S0tHS43W5yyy239EiSVLjjjjtYNwWmaYKiKMDjUkpBkiQuF+Diahsed2xsDPr6+mDjxo3MbiaTgbq6Orj99tuZXVEUwe12c825WCwCAHC5hmGAqqpcLiEEZFm+5PUdHR2F8+fPwxe+8AVmN5lMQnd3N9x6663Mbi6XA5/PB5s3b2Z2ZVmGt956q+zHMwf3Jz/5yWlN0877fD4RAMDj8egul4t6vV7WTYFhGFBVVQU8LqW0ItflcnG5Ho+He9xK3Er2FQBwP19d17nHJYT8W9VIVdWK9pXdXv4hJ+bg1tfXKwCgsHoIgkwceDoIQSwIBhdBLAgGF0EsCAYXQSwIBhdBLAgGF0EsCAYXQSwIBhdBLAgGF0EsCHNw29vbZ+zfv39VKpViv/obQZAJgfmSx2AwqD7zzDOLv/zlLxe/853v9BuG4SCEcK1aMU0TTNPkciml3C7AxWtDL/WcK3ENw6hoXN7nW8m4hJCK6lvJnK1YX5Z7sDEHd82aNUpXV1dClmXXh8v6VnV0dNSwbgfgP5d98SwJBLi4NI9nSSAAcC8nzOVyEA6HIZPJMLuFQgEGBgZAkiRmt1gsQl9fH2iaxuzqug7d3d3czaiOHz/OdAF8CUopHD9+nGtJIAB/fbPZLMRiMUgkEsyuKIoQDAYhl8sxu4qiwNmzZ0FR2C/lZ60r8x7dt2/fFQ6Ho37t2rUBt9tNtmzZ8oFhGOKdd7K3zDUMA+x2O9x1113Mbukdmdc1TZPLjcfj0NPTw7XsK51OQ3t7O9eyL1EUob6+HlpbW5ldRVHA6/XCtm3bmF1d18HhcMD27duZ3dIbxaWuUTQahUAgwLX0MpFIQFdXF9fSS0EQoK2tDbZs2cLsFgoFeP3118t+PHNwH3744UEAmJA7X+B9lRGED7yvMoJYEDwdhCAWBIOLIBYEewchiAXB37gIYkHwqzKCWBAMLoJYEAwuglgQDC6CWBA8qowgFoQ5uLlczv7QQw/9n2g0Or3SwfGoMoLwwXyt8p49exa9+OKL8+65556exsbGzNDQ0Ox4PO46d+4c8+CEEIjH48DjUkq5XYCLPYB43FQqBdFolMvN5XIwOjrK5RYKBYjFYlyuqqrc+8o0TcvVaHx8nLtGmUyG2xVFkbu+iqIwrRBiDq4gCHMXLVo0c3h4uHHJkiURQRB8hULBwbPMjRACkiRxLZEDuPhi5nV5xxUEAURR5HLz+Ty3qygK95w1TeN2TdPkdimlFdWI162kRpW4hUKBe18Vi0WmZZfMwd2xY8fha6+9tn/p0qWSy+Wiq1atCgaDweL111/PuikwDAOGh4eBx6WUwuDgILc7MDDA5cbjcXA6nVxuOp0GTdO4XFEUIZvNcrmKokA8HudydV2HkZERLpcQAkNDQ5e8RtFoFKqrq7ncRCIBlFIuVxAEkCSJyy0UCpBKpcp+PNcK57Vr147zeAiCTAx4OghBLAieDkIQC4KLDBDEguBXZQSxIBhcBLEgGFwEsSAYXASxIHhUGUEsCB5VRhALwhzckZGRxgMHDqyRJMk7GRNCEOTTYb7k8b333ruqq6vLs3btWgoA8GHTL5uu68yDl5ok8bilFhU8bmlsHrfUBAvdT6fU9Iu3RoQQSz3fSt1Jbfo1d+7cU0ePHl35zDPPtDz44IOnP2z65ef5vVpqKMXTFIpSCp2dnVBVVcXsAgB88MEH4Ha7mT1BECAcDkM2m2V2JUmCgYEBKBQKzG6xWITe3l6uF4Wu63DixAlmD+DiGxxvY7ZSjXgbs/HWN5PJQCwWY7pov0Q+n4dgMAj5fJ7ZlWUZzp49C8VikdlVVZXp8cyJOX36dJ3P55u2du3a/olo+mWz2bgaSpWafvG6hmFwuVPZ9Kuuro676ZfH45mSpl+U0kteI2z69T/wve99bwgAhlg9BEEmDjyPiyAWBIOLIBYEg4sgFgSDiyAWBIOLIBYEg4sgFgSDiyAWBIOLIBYEg4sgFgSDiyAWhDm477777swf/OAHm1999dVZkzEhBEE+HeZrlX/zm9+ssdls3muuucZmmqbtwoULc2KxWFUgEGAe3DRNiMViwONSSrldgIuLBXjcZDIJ4XCYyxUEASKRCJdbKBQ+unieFVVVYXR0lMs1DIN7PxNCKqovb43Gxsa4a5TJZLhrJIoid40mvenXvHnzVAAonjlz5qrW1taEKIpeRVHsuVyOdVNACAFZloHHpZSCoihcLgBwu6Iocs+5EleWZe45a5rG7ZqmiTUqE0mSuN1L0fSr48SJE1c1NzcHnE4nveaaa4YvXLhQXLNmDeumwDAMGBoaAh631BSK1z137hyXG4/HweFwcLnpdBqKxSKXW+ogx+MqigKjo6Ncrq7rMDw8zOUSQmBwcPCS1ygajYLX6+VyE4kEmKbJ5ZY6/fG4hUIBEolE2Y9nDu7ChQvFhQsXHmf1EASZOPCoMoJYEAwuglgQDC6CWBAMLoJYEK6O9BMFdjKYPCiloOs6JBIJiEajEA6HIRgMQmNjI3g8Htz3FmdKg4udDCae0t0Ru7u7YWBgAGbNmgWzZ8+GTZs2QSgUgmPHjsGsWbNg3bp14PV6McAWZUqDi0w8xWIR9u/fD01NTfCVr3wFqqqqPgpn6Sby58+fh927d8MXv/hFmD59OobXguBv3M8QpmnCwYMHoaWlBa6//npwu93/JZQ2mw2cTie0tLTALbfcAgcOHGC+ETfyrwEG9zNEMBiEqqoqaGlp+aefojabDWbPng3Lly+Hzs5O/MliQZiDK4qiK5vN1iiKwtdXApkUKKXQ09MDq1evLuurr81mg5aWFgiHw9y9fZCpgzm4vb29s++///6v/O1vf1s2GRNC+NB1HYrFIjQ0NJTtuFwu8Pv9IEnSJM4MmQyYD07dcMMN0V/84hfC+vXrh1RVtR84cOCaDz74wM/TQIsQAl1dXeD1snfspJRCV1cX+Hw+ZhcA4MSJE+D3+5k9QRAgFApxvdglSYJz584xLd8qoSgK9PT0fOK/q6oKoigyHWiy2WzgcDhg9+7dnxh40zThgw8+AI/HwzznUo2qq6uZXQD+GqXTaRgdHeVapZPP5+HChQugKAqzK8synDlzBkzTZHZVVWU63sAc3IGBAb/H4zGWLFmSBwDYsGFDIJ1Oy5s2bWLdFJimCblcDnhcSilkMhkuFwAglUpxuePj49DX1wc333wzs5vJZMDr9XKNW3qj+CRX0zSmplEA/3mUedOmTZ8YXMMwQBRFrjkTQiCbzXLXl7dGo6OjMDAwADfddBOzm0qloLa2lmvcXC4HTqeTy5VlGdra2sp+PHNwL7vsMvm3v/3tm6X/9/v9RY/HQ2pqalg3BYZhgMfjAR6XUjolriRJ4PV6uVxN07hdAPinbuk55XI5qKurK2t7hmGAJEnQ2Nj4ie0sdV3n3leEkCmpkd/v597PiqJwu6Zpcrt2ux3s9vJ/uTL/xvX5fGTWrFnsDUD/B/D84cRhs9lg2bJl0NXVVdZRYkopDA4OQmNjI3f/WmTqmNLTQXgaYmJZvHgxCIIAwWDwn+5bSilks1no6uqCtWvX4huoBZnS4OILZmJxOp1w2223wfHjx+H06dNgmuZ/CXCpGXgoFIK9e/fCxo0buQ8cIVMLXqv8GcPn80Frayu0t7fDiy++CPPmzYPGxkZwOByQTCYhFAqB2+2GLVu2QG1tLb55WhS8Vvkzhs1mA7fbDTfddBPIsvzRyqCOjg644447YNOmTVBbW8t0IAT51wOD+xnFZrOBz+eDJUuWwLx58yCRSMDy5cunelrIBIFvuwhiQTC4CGJBMLgIYkEwuAhiQZiDGw6Ha99+++2rBwYGKj4BiKciEIQP5qPKP/rRj649cuTI4vvvv9/+/e9//xQhxFY6sc8KIQR4XUopt1vyL/W4lbhTta8qGdeKc660voSQiuZcLszBNU2TTp8+Pe10Ot2qqtpfffXV1e3t7TW8k+3o6GD2AC7upGPHjnGfjzx27Bg4HA5mTxAECIfDkEwmmV1JkmBwcJC7KVRfXx/XcjNN0+DkyZNgGAaza5om910yCCFw7Ngxrm9WlFLo6OjgqlEmk4F4PA5jY2PMbj6fh2AwCJlMhtmVZRnOnj0LhUKB2VVVlem1zBzc9evXZyRJmlNfX59wu93kS1/6UichRLzzzjtZNwWGYYDD4YC77rqL2S29kO6++24ulxDC5cbjcejp6YFbb72V2U2n09De3g6bN29mdkVRhIaGBmhtbWV2FUUBn88H27ZtY3Z1XQeXywXbt29ndgkhYLfbuWtEKeVyS60uN27cyOwmEgno6uqC22+/ndkVBAHa2tpgy5YtzG6hUGBalskc3Hvvvff0vffee5rVQxBk4sCjyghiQTC4CGJBMLgIYkEwuAhiQTC4CGJBMLgIYkEwuAhiQTC4CGJBMLgIYkHwLo8IYkGYg7tv374rd+zYsbm3t7e+0sHxLo8Iwgfztco7d+5sjkaj9KWXXlq5YsWKd8Lh8IxkMukKBoPMg5umCYlEAnhcSikkk0kuFwC43WQyCWNjY1yuIAgwPj7O5RYKBe59paoqt2sYhuVqNDY2xl2jTCbD7YqiyF1fRVGY2p0yB3fHjh0n/vCHP3zB5XIldV23xWKx+nQ67YxEIqybAkIIZDIZ4HFLTb94XADgdjOZDCSTSS43n89zu7IsQyqV4nI1TYN0Os3lmqZZUY14xwXgr1EymeTez4IgcO9nSZK43WKxyLTskjm4M2bMcNx0002xjRs3nna5XHTNmjWDkUhEufHGG1k3BYZhQDQaBR6XUgojIyPc7vDwMJcbj8fB6/Vyuel0GiilXK4oilAoFLhcRVEglUpxubquQywW43IJIRAOhy95jaLRKEybNo3LTSQS4HQ6uVxBEEDTNC530pf1NTc3jzc3N4+zegiCTBx4OghBLAgGF0EsCJ7HRRALgv1xEcSC4CcuglgQ/MRFEAuCB6cQxIJgcBHEguBvXASxIGUFV1VVR1tb29Xj4+O+t99+e2lnZ+fsiRgcf+MiCB9lXfJomib87Gc/W/TAAw/4du3aNUtV1aqXX375ZbvdTgkhdkopmKbJPHipQRKPC3Ax+JfarWTOlbqVzHkqnm+p1ctU1Miq9S2XsoJbXV1tXnnllXEA8KxevTo1MjJyGQDAh02/rq206RdvU6hKmn7xNpTK5XIwMjICqVSK2S01/crn88xuJU2/dF2H7u7uipp+8TBVNaq06dfw8DBks1lmV5ZlCAQC/zpNv4rFosvn81155syZ3PDwsGfOnDkxu91O3W43xaZf5TOVTb+qq6unpOmXzWa75DXCpl8f4vF49F/96lfPM88GQZBJAY8qI4gFwSunEMSC4AUYCGJBMLgIYkEwuAhiQTC4CGJBMLgIYkEwuAhiQTC4CGJBMLgIYkHKCq4oiq5HHnlka09PT9N3v/vdLx09enTmZE8MQZBPpqxrlWtqavRIJJKglHoaGxtT586d869bty4JADAyMjIzkUi4hoaGmAcvNf3icSml3C7Axf4yPG4ymYR4PM7lCoIAY2NjXG6hUIDx8XEut9T0i8ctNf3icQkhU1KjsbEx7hqVVhbxuKIoctd30pp+ud1uV7FY9NTV1bk0TfMAAOi6bhsbG5uWzWado6OjzJMlhEA2mwUel1LK7QIAt5vJZCCdTnO5+Xye25VlGTKZDJeraRq3W2r6ZaUaJZNJSKVSXK4gCNw1kiSJe19NStMvVVWdGzduJKOjo/66ujrS0NDgAwBwuVz0uuuuuxAOh5UNGzYwT9YwDIhEIsDjUkohFApxu8FgkMuNx+Pg8Xi43HQ6DYQQLlcURZAkictVFAWSySSXq+s6jI6OcrmEEBgZGbnkNYpGo1BbW8vlJhIJcDgcXK4gCKCqKpc7Kcv63G638dWvfvUw82wQBJkU8KgyglgQDC6CWBBcSI8gFgQX0iOIBcGvyghiQTC4CGJBMLgIYkEwuAhiQTC4CGJByg5uoVDwGIZhlyTJI8sye18IBEEmjLKCm8vlqlpbW7cfO3as+ZFHHvlfO3bsuGWyJ4YgyCdT1rXK06ZN0+bNmxeYMWOG1traGnjqqadaAAA0TbMfOnRoxcmTJ31+v595cEIInDx5Empra5ldSimcPHkS6urqmF0AgFOnTkF9fT2zl81mYXh4GFRVZXZFUYRAIMB1/lpRFDh16hS4XC5mV9M06O7uhurqambXNE3o7u6GmpoaZrdUo2nTpjG7APw1SqfTEI1GQZZlZjeXy8Hg4CBXg7RCoQB9fX1cTc5UVWVqFlZWcDVNc0iSNO/dd9+tefbZZ1c8+OCDrwMAVFVVkdWrVw/FYjFl/fr1zJM1TRPGx8eBx6WUQiwW43IBLq4g4XETiQT4fD4uN5vNgt1u53IlSYJiscjlqqoKuVyOyzUMA5LJJJdLCIF4PM5do9HRUS43FotBfX09l5tKpaCqqorLzefzYJomlyvLMrz//vtlP76s4DocDvLEE08cMU2Tbtq0qd/j8Xz01lBXV1fw+XykoaGBebKGYYDP5wMel1I6Ja6qqlBTU8M9rt/v53JdLhe3qygK9/PVdZ3bJYRUVKPq6mouV5Zl7n1lGAZ3fe12O/e4brebqaVoucGlCxYsiDHPBkGQSQFPByGIBcHgIogFweAiiAXB4CKIBcHgIogFweAiiAXB4CKIBcHgIogFweAiiAUpK7iaptk7OzuXplIpf2dn55Kenp4ZEzE43uURQfgo65JHXddtjz766Iof//jH5rlz5+r2799/zZ49e/4OAEAptX34X+bBKaUf/fG4vOP+923weJd6XCs+30rq+/Ft8DpTUd9Kn2+5lBVcn89nLl68OFpfX280NzenbTbbAgAAVVXtr7zyynXvv/9+Dc8yKEIIdHR0cBenvb2d2fu4y/OJLwgCRCIRSCQSzK4kSTA4OAjZbJbZLRaL0NfXx7T0q4SmaXDy5EnQNI3ZNU0TOjs7gRDC7Jbqy0MlNSp13IvF2C+vz+fzMDw8DOl0mtmVZRkCgQCIosjsqqrKtBywrOAWi0WX0+ls3rt37yy73W4uWrQoDQDgdrtJa2vrcUqpeOeddzJP1jAMcDqdcNdddzG7lFKw2Wzw9a9/ncsFAC43Ho9DT08P3HrrrcxuOp2G9vZ22Lx5M7MriiIcPHgQWltbmV1FUcDv98O2bduYXV3XoaqqCrZv387sEkLA4XDA3XffzexWUqNoNAqBQAA2btzI7CYSCejq6oLbb7+d2RUEAdra2mDLli3M7qQ0/fJ4PPqTTz7519LOdDgceCdzBJlCyu6Pa7fbMawI8i8C9g5CEAuCvYMQxILgBRgIYkEwuAhiQfA3LoJYEPyNiyAWBL8qI4gFweAiiAXB4CKIBSkruJIkuR5//PH/Oz4+3nDo0KGFv/vd71ZN9sQQBPlkyrrk0e12G/39/bkzZ87MevLJJ1eYpqk+8MAD3QAA8Xi8LpvN2qLRKPPSE9M0IZPJ2KLRKPNRKkopfDgu1xEuXjeRSEAqleJyBUGAdDrN5RYKBe59paoqt2sYRkU14nUB+GsUj8e5a5ROp7ldURS566soCmFZYVdWcF0uF/X7/XD48OHLpk+f3jQwMODIZDL2mpoaeuHChdmjo6N/CwQCJ1knSwixx2KxlYFA4ASrSym1xWKxzwcCgS5WFwAgFotdy+NmMpnaSCTSFAgEzrG6oij6IpHIFYFA4AyrqyiKJxqNLgkEAj2srqZprtHR0WWBQOAUq2uapiMWi/1HIBDoZnUrrBG3m0ql6sLh8KxAIDDA6gqC4I9EIvMCgcBZVrdQKHij0eiiQCDQx+oWi8UqTdOqyhY+vtj5k/4URXE+++yzt+7Zs+e6fD7vfe211xaW/k3TNMfrr7++vJzt/Pc/wzBs+/btu5rHpZTCnj17/uNSu8lk0n/06NFFPK4gCN62trYlPK4kSVWHDh1q4XGLxaLzjTfe4KqRruv2/fv3r+Ddz3v37r3k9R0bG6vt6OhYyOOm0+nq9957bzGPm8/n3W+//fbneFxZll1vvvnmVeU+vqzfuB6Px/jmN7/55ubNmztramqUO+64I1j6t7a2tsuTyaSvUCgwd6nv7Oy8ora2lusqjN27d18pCEJ1Npst/13qQ1599dXLs9msP5VKeVhdr9dbGBoaqs3lcszNZhOJxMx4PH5ZPp/3sbrJZNIejUbrjx49OpPV7e/vr0ulUnMOHz68mNV99913r0gkErWhUIh5Px85cmRBIpHw5nK5slehFYtF5wsvvLBheHi4PpvNVu3du3dhua6mafadO3euKxQK7v7+/sbTp09PL9c1DMP297///fpMJjMtHA437N69e0m5LiHEtnv37uvi8fhMURS9u3btWqZpWtmv69dee21VNBqd09LSEnrhhRdWluNUdFQ5k8k4fvnLX1578ODB2mPHjjWx+uPj4+Irr7xS9g76OA0NDcl33nmn/sCBA7NYXVmWc7///e8XtbW1zWF133nnnbkPP/zw/04kEszdmk+cOHHle++959M0jfl2Ek899dTKffv2zQmFQszu4sWLhd7eXiMYDDJ1Hy8UCrZHH310zU9/+tMVoVCIqcO0LMv2H/7wh9c9/fTTsw4ePFh2+Gw2G3nllVeqe3p6lnV0dEx//vnnrzVNs6wQ2Gw2unfv3qpwODw7GAxWvfXWW7PLHdfpdNJDhw7ZBwcHmxoaGoLPPffcKlmWy3rDsdvt9MiRI6Svr2+u0+nM//GPf1xbLBbLfrPq7OzUOzo6Fv76179e+fjjj1+fSqU+9flWFFxKqa2qqorOnDlTNk2TuVX6/PnzRZvNZvKMHQ6HfYVCwX3zzTcz30Nm2bJlxRUrVoyKouhmdY8cOfL5yy67rDYcDpf9oigxe/bsky6XK/+nP/1pOasrSVJNU1OT2tXV9XlW1+l0mv39/TNvvPHG8yyeaZo2l8ulzZgxI65pGtM3jOrqanLfffedcrvdjbqul/0idrvdpKmpKUcpdc6dO1dpbGwse0yXy0Xnzp2b83g85LrrrsuzzBcA4PLLLxccDgft6+tbNG/evAvV1dVlHy2aP3++YLfbobm5Of1h/+iyszV//vxcT09PdW9v79V+v78pHo9/6jfBioI7ffp0Y9OmTUNVVVX1K1eujLL6+Xx+QV1dXX0oFGL+9BoeHl52ww032PL5PPPX3cOHDzfPmTPHu379+nFW9+c///nehx56aFdLS8sIqzswMNAwbdq0+ttuuy3E6n7729/urq+vd65Zs2aQ1c1kMo5169aF5s+fL7N4tbW15Gtf+1pk1apVZNasWUnWcWfPnm1fv379yIYNG4bKdVRVdc6dO3duKBRyS5LkX7BgwVC5d1zRdd0xZ86cy7u7uxcMDQ1d7vf7FxBCyvq0NgzDPn369PlnzpxZGI1Gr7rhhhtchmGU5RJCbH6//4qhoaHL33jjjdVr166NuVwuvRwXAMDtdl8xd+7cmp07dz73rW99a+/8+fOLn+bYKMXrhRHEauCVUwhiQTC4CGJBMLgIYkEwuAhiQTC4CGJBMLgIYkEwuAhiQTC4CGJBMLgIYkH+PwdzHyydJxCfAAAAAElFTkSuQmCC",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -482,14 +471,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"As the game has just started, I will place my piece near Alice's to prevent her from forming a line and to start building my own line.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        8\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -510,10 +499,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -524,14 +511,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"My opponent placed his piece next to mine. I should block him and also try to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        6\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -552,10 +539,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -566,14 +551,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"I see Alice is trying to create a line horizontally. I need to block her next move otherwise she will have a chance to win. I will place my piece at [7, 5].\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        5\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -594,10 +579,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -608,14 +591,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [7, 9] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        9\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -636,10 +619,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -650,14 +631,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"Alice is trying to create a line horizontally again. I need to block her next move otherwise she will have a chance to win. I will place my piece at [7, 10].\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        10\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -678,10 +659,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAmkUlEQVR4nO2deXBU15m3f7cXdUvd2tnEHrBYBY5NzGKWfHaCwcHGYRBJGLCdZLJ4xi5PkkrNVFLJZGbK+eIJM6nJJIHKVBwnBmMw2A6LwRBjGRsQQgiQhGjQ0pJarW71vu/33vP9YZovHmPoc8QyF79PlcqF1c89R/ftXy/33nNfiTEGgiC0he52T4AgCH4ouAShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUG4g3vhwoWKrVu3LhoaGiq5GRMiCOL6SLwXYKxfv355ZWWlPHbs2OSPfvSjpkwmY2CM6W/S/AjiE4PBYJANBoNS0GN5Ny7LsmHGjBler9dbnslkdPv27fvM2bNnn6urq6vh3Zaqqjh//jzmzp3Lq4Ixhra2Ntx9993cLgC0trYKufF4HB6PB1OnTuV2U6kUBgYGMG3aNG43k8nAbrdj5syZ3K4sy7h06RJmz57N7aqqio6ODsyZM4fbZYyhvb39ltc3Go3C7/djypQp3G4ikYDL5UJtbS23m06n0dfXhxkzZnC7uVyOFRcX/99169ZtL0hgjHH9vPrqq5O/8Y1vfKm5uXkkYwyZTMa4bdu2NiZALpdjL730kojKVFVlf/jDH4TdF198Uch1uVzs4MGDQq7f72d79uwRcqPRKNu1a5eQm0wm2fbt24XcbDbLtm7dKuQqisL++Mc/CrnDqdHAwAA7fPiwkOvxeNj+/fuF3FAoxF5//XUhNx6Pyzt37nyKFZhD7nfcdevW9a1bt66P1yMI4sZBR5UJQoPc1uBKknQ7hycIzXJbg8toSSFBCEEflQlCg1BwCUKDUHAJQoNQcAlCg1BwCUKDUHAJQoPQeVyC0CDcwT127NiEH/3oRxtuxOB0HpcgxOC+VnnSpElum82mqqoqqaqKCxcujHc4HKZz585xD66qKhwOB0RcxpiwCwADAwNCrt/vh91uF3IjkQh6e3uF3GQyib6+PiE3m82iv79fyFUURdi9XTXyeDzC+yoUCgnXNx6PC9c3nU4jk8kU/Hju4EqSZMxkMmWRSMRcWlqaVlVVYoxBVVXeTUFV1Ss/IoiO+5dji4wpOu5wXFVVNecO5+/N+5+k/cwDd3B1Op35W9/6VmcgECiprKxM1dXVDdhstsy9997LuynIsgybzQYRl11e6ynqtrW1CblutxuKogi5gUAA0WhUyI3FYhgaGhJyU6kUent7hdxcLodLly4Jufm1vLe6Rk6nE3q9Xsj1er1Ip9NCbjgcRiAQEHITiQScTmfBj+cO7tixY0OrV69u4PUIgrhx0OkggtAgFFyC0CAUXILQIBRcgtAgdOUUQWgQugMGQWgQ+qhMEBqEgksQGoSCSxAahDu4oVDI5PV6y2VZpiNLBHGb4L7k8Xe/+12NzWa7b9myZee++tWvdt2MSREEcW24g/vEE08MPPPMM7PLysqS+aZfJ0+etOr1/A37VFXFyZMnYTQauV3GGJqammAymbhdAGhqaoLZbOb2wuEw+vv7EQ6Hud14PI5Lly4hmUxyu6lUCm1tbZBlmdvNZrM4ffq00Ok3RVHQ2NgIg4H7qXKlvkVFRdxuvr4iNQoEAnC5XAgEAtxuNBpFT08PYrEYt5tMJtHR0cG1PC9PJpPhqi13NX7605/OLy4unjxr1qxGk8mkPvTQQ63xeDzx2GOP8W4KiqIglUpBxGWMIR6PC7nAB6ttRNyhoSG0t7dj+fLl3G4wGERFRQVWrVrF7cZiMZhMJqE5p9NpABByZVlGJpMRclVVRTKZvOX1HRwcxKVLl/Dggw9yuz6fDy0tLVi5ciW3G4lEYLFYsHr1am43mUziz3/+c8GP5w7uv/zLv5zLZrOXLBZLDADMZnPOaDSy4uJi3k1BlmUUFRVBxGWMDcs1Go1CrtlsFh53OO5w9hUA4b83l8sJj6uq6ieqRplMZlj7Sqcr/JATd3ArKytTAFK8HkEQNw46HUQQGoSCSxAahIJLEBqEgksQGoSCSxAahIJLEBqEgksQGoSCSxAahIJLEBqEO7jHjx8fsX///nl+v5//6m+CIG4I3Jc82u32zAsvvDDtS1/6Uvrv/u7vOmRZ1quqKrRqRVEUKIoi5DLGhF3gg2tDb/Wch+PKsjyscUX/3uGMq6rqsOo7nDlrsb4892DjDu7ChQtTzc3N3mQyaby8rG9eY2NjKe92gP+/7EtkSSDwwdI8kSWBAISXE0YiETgcDgSDQW43kUigs7MT8Xic202n02hvb0c2m+V2c7kcWlpahJtRnTx5kusC+DyMMZw8eVJoSSAgXt9QKASXywWv18vtxmIx2O12RCIRbjeVSuHChQtIpfgv5eetK/ce3bdv36f0en3lokWLbCaTSV2zZs0pWZZjGzbwt8yVZRk6nQ6PP/44t5t/RRZ1FUURct1uN1pbW4WWfQUCARw/flxo2VcsFkNlZSXq6+u53VQqheLiYqxfv57bzeVy0Ov12LhxI7ebf6G41TVyOp2w2WxCSy+9Xi+am5uFll6Gw2E0NDRgzZo13G4ikcCbb75Z8OO5g/u9732vC8ANufMF3VeZIMSg+yoThAah00EEoUEouAShQah3EEFoEPqOSxAahD4qE4QGoeAShAah4BKEBqHgEoQGoaPKBKFBuIMbiUR03/nOdx51Op3Vwx2cjioThBjc1yrv2bPnrldeeWXiN77xjdaamppgT0/PaLfbbbx48SL34Kqqwu12Q8RljAm7wAc9gERcv98Pp9Mp5EYiEQwODgq5iUQCLpdLyM1kMsL7SlEUzdXI4/EI1ygYDAq7sVhMuL6pVIprhRB3cMPh8Pi77rprZG9vb8306dMHwuGwJZFI6EWWuamqing8LrREDvjgySzqio4bDocRi8WE3Gg0KuymUinhOWezWWFXURRhlzE2rBqJusOp0XDcRCIhvK/S6TTXskvu4D777LPv3HfffR0zZsyIG41GNm/ePLvdbk/ff//9vJuCLMvo7e2FiMsYQ1dXl7Db2dkp5LrdbhgMBiE3EAggm80KubFYDKFQSMhNpVJwu91Cbi6XQ39/v5Crqip6enpueY2cTidKSkqEXK/XC8aYkBsOhxGPx4XcRCIBv99f8OOFVjgvWrTII+IRBHFjoNNBBKFB6HQQQWgQWmRAEBqEPioThAah4BKEBqHgEoQGoeAShAaho8oEoUHoqDJBaBDu4Pb399ccPHhwYTweL74ZEyII4vpwX/L43nvvzW5ubjYvWrSIAcDlpl9SLpfjHjzfJEnEzbeoEHHzY4u4+SZY5F6ffNMv0Rqpqqqpv3e47k1t+jV+/Pizx44du/eFF16Y9cwzz5y73PTLKvJ9Nd9QSqQpFGMMTU1NKCoq4nYB4NSpUzCZTNxeOByGw+FAKBTiduPxODo7O5FIJLjddDqNtrY2oSdFLpfD6dOnuT3ggxc40cZs+RqJNmYTrW8wGITL5eK6aD9PNBqF3W5HNBrldpPJJC5cuIB0Os3tZjIZrsdzJ+bcuXMVFoulfNGiRR03oumXJElCDaXyTb9EXVmWhdzb2fSroqJCuOmX2Wy+LU2/GGO3vEbU9OsqfPe73+0B0MPrEQRx46DzuAShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUEouAShQbiD++677478x3/8x9VvvPHGqJsxIYIgrg/3tcr/9V//tVCSpOJ77rlHUhRF6u7uHuNyuYpsNhv34IqiwOVyQcRljAm7wAeLBURcn88Hh8Mh5IbDYQwMDAi5iUTiysXzvGQyGQwODgq5siwL72dVVYdVX9EaDQ0NCdcoGAwK1ygWiwnX6KY3/Zo4cWIGQPr8+fOz6+vrvbFYrDiVSukikQjvpqCqKpLJJERcxhhSqZSQC0DYjcViwnMejptMJoXnnM1mhV1FUahGBRKPx4XdW9H0q/H06dOza2trbQaDgd1zzz293d3d6YULF/JuCrIso6enByJuvimUqHvx4kUh1+12Q6/XC7mBQADpdFrIzXeQE3FTqRQGBweF3Fwuh97eXiFXVVV0dXXd8ho5nU4UFxcLuV6vF4qiCLn5Tn8ibiKRgNfrLfjx3MGdMmVKbMqUKSd5PYIgbhx0VJkgNAgFlyA0CAWXIDQIBZcgNIhQR/obxSehkwFjDLlcDi6XC5cuXcLFixcxffp0TJgwAcXFxdfcB3nX6/Wiv78f/f39sNvtqKmpgdlsLth1Op1wOBzcrsvlgtPpRFdXFy5dusQ9Z4/Hg4GBAaFxXS4Xuru7cenSJUycOPG67ieN2xrcO7mTQf6+z++++y42b96MpqYm+Hw+KIqCTZs2Yfbs2fj617+O+vr6j4Qh754+fRqdnZ0YNWoURo8ejc997nPo6+vDiRMnMHLkSCxduvSqrizLaGlp+ZC7YsWKK+6oUaOwZMmSq4Ywl8tdmfOpU6euzHnz5s2oq6vD3/zN3+Cv/uqvCh531apV1x037zY0NGDLli0fGve3v/3tNcf9xMIYG9ZPJpMxbtu2rY0JkMvl2EsvvSSiMlVV2R/+8Adh98UXXxRyXS4XO3jw4HUfl0wm2T/90z+x8vJyBuCqPyaTiW3cuJENDQ19xH311VfZsWPHWDqdZqqqfmjuuVyOdXR0sD/+8Y/M5/N96PeFuOfPn/9Y9yc/+QkrKyu75pw3bNjAPB7PR9ydO3cWNK7f77/ye1VVWSKRYD/+8Y+vO+7jjz/OvF7vh7Z9NQYGBtjhw4evW6Or4fF42P79+4XcUCjEXn/9dSE3Ho/LO3fufIoVmLvb+o57p5J/p3j++eeveRlbJpPB9u3bYTAYsHnzZhQXF0NRFBw+fBizZs3CrFmzPvLuIkkSDAYDZs6ciaqqKhw8eBDr1q2D2WyGoig4dOjQdd1Zs2ahurr6I+6WLVvw/PPPX/Pm3JlMBq+88gqKioqwefNmmM1mqKqKt956C3V1dZg5c+Y1x62qqsKBAweujMsYw5YtW/Dzn//8uuO+/PLLMJlM+NWvfgWz2XytEtzx0MGpGwxjDJcuXcKmTZsKuvZUVVXs3LkTBw4cAGMMvb29V57k1/pIKEkSRo8ejTlz5uDkyZNgjMFut6OoqIjLbWpq+tCcC7mjvqqq2LFjB9566y0wxtDd3Q2z2YwZM2Zcd9wxY8Zg9uzZOHXqFBhjsNlsXONu27YNhw8fvqO/ZhUCd3BjsZgxFAqVplIpsb4SnwDeeOMNuN3ugh+fSqXw0ksvQVEUtLa2Yv78+QV9j5MkCbNmzYLD4YAsy8JuLpfD7t274fF4uOa8detWZLNZtLa2YsGCBdDprv90kiQJdXV16O/vRzabxWuvvcZ1qV86nca2bdsgy3LBzp0Id3Db2tpGP/XUU19++eWX627GhLSOqqo4fvw49zvChQsX4PV6kUwmUV1dXbBnNBpRVlYGv9+PdDqNqqoqLtdqtSIQCOD06dPccz579iyCwSByuRzKy8u5xi0tLUUoFMKxY8e4xz1z5ozwwoU7Be7vuIsXL3b+/Oc/Dy9durQnk8noDh48eM+pU6esIg20VFVFc3Mziov5O3YyxtDc3AyLxcLtAsDp06dhtVq5vXA4jL6+PsTj8av+PpfLoa+vj3u7Ho8HO3bsgMlkKuidK48kSTAajdi1axeMRiPXEVdJkqDX67Fr1y7hZYq7d++GyWTiPtIrSRJef/11dHZ2co/r8Xjw+uuvf+yLVCAQwODgoFC4o9Eouru7kUqluN1kMonz589DURRuN5PJcDX+4g5uZ2en1Ww2y9OnT48CwLJly2yBQCC5YsUK3k1BURREIhGIuIwxBINBIRcA/H6/kOvxeNDe3o7Pf/7zV/29oih44YUXuINQXV2NVatWobGxEaqqFhxedvnc58qVK3HyJN/aD3b5tNPDDz+MAwcOoLu7m8svLS3FihUrrnxf5R37oYcews6dO9Hf38/lVldXY/ny5RgxYsRVfz84OIjOzk488MADXNsFPnhelJWVCT03IpEIDAaDkJtMJtHQ0FDw47mDO27cuOSvfvWrt/L/tlqtabPZrJaWlvJuCrIsw2w2Q8RljN0WNx6Po7i4+GNdxhgWLVqEI0eOcG13+vTpmDJlCtra2hAOhwv+yCvLMqLRKCZPnoxz584hEomgoqKiYDcej2PixImYN28eDh06xDXnOXPmYMKECTh79izi8XjBH5cVRUE8Hse4ceOwZMkSvP/++1zB//SnP41x48Z9bAtOq9V6zRpdi1QqJewqiiLs6nQ6rk9a3N9xLRaLOmrUKP4GoFfhTj2RvnbtWowcObLgxxcVFWHjxo0wGAyoq6tDS0tLwU/k7u5u1NTUwGg0oq6uDs3NzQW5jDF0dXWhpqYGRUVFqK+v5/puXVRUhA0bNsBsNnOPe/HixStXUq1duxaVlZVc465fv1645+6dwm09HXQnHtLPHzV95plnCmrYLUkSHn30UTz22GOQJAm1tbUIh8Ow2+3X3D/5rwqnTp3C/fffD0mSMG3atILdUCiE5uZmLFq0CJIkYc6cOXj66acLCoQkSVi9ejUeffRRSJKE6dOnIxgMore3t6BxW1pasHDhQkiShLlz53KN+8UvfhGrVq26Y1/0C+W2BvdO3fkGgwHf/e538dRTT13zwJter8cXvvAF/OIXv7jy8cpgMODhhx/GyZMnce7cOciy/KEwsMsNvfv7+7F3714sX74cJSUlV3UVRbmq29fXd1X3+9//Pv72b//2mhc35Mf4xS9+ceXAYP7/nThxAq2trR87bm9vL/bu3YuHHnroyrh6vR7f//738e1vf/u64z7yyCP4j//4D+EDkncSdK3yTcJqtWLTpk1YunQptmzZgtbWVkQiETDGYLVaMWXKFDz55JP42te+9pHvhhaLBfX19Th+/Dh27NiBiRMnoqamBnq9Hj6fD319fTCZTFizZg3Kyso+9AL4l+4rr7zC5VqtVvzbv/0bFi9ejN/85jdob2//0JynTp2Kr371q3jyySdRXl7+Efda4/b29qK4uPgj40qShNLSUmzatAmLFy/Gli1buMb9pEKXPN4kJEmC2WzGunXrsGrVKtjtdpw/fx4dHR1YvXo1amtrUV5eftUDEpIkwWQy4YEHHkAymYTT6URvby9aWlrwyCOPYMWKFSgrKyvYdTgcaGxsxGOPPXZdNz/nL3zhC7Db7bDb7Xjvvffwla985bpzNpvNePDBB6+M63K50NDQgLVr12LlypXXHffLX/4yHnnkEdjtdvT19eHIkSPYsGHDNcf9pELBvclIkgSLxYI5c+Zg7NixsFgsmD9/Ppc7ffp0jB07FtFoFHPmzOF2J06cCK/Xy+VarVbMnTsXM2fORDweF5pzbW0t+vv7hcadM2cOgsFgweN+0qCXMILQIBRcgtAgFFyC0CAUXILQINzBdTgcZW+//fbdnZ2dJcMdnA7rE4QY3EeVf/KTn9x39OjRaU899ZTuH/7hH86qqirlT7DzoqoqRN38LTxE3Lx/q8cdjnu79tVwxtXinIdbX1VVhzXnQuEOrqIorLq6OmAwGEyZTEb3xhtvzD9+/Hip6GQbGxu5PeCDnXTixAnhc3snTpyAXq/n9sLhMBwOB3w+H7cbj8fR1dUl3BSqvb1daLlZNpvFmTNnhBafK4py5S4ZvKiqihMnTgh9smKMobGxUahGwWAQbrcbQ0ND3G40GoXdbkcwGOR2k8kkLly4gEQiwe1mMhmu5zJ3cJcuXRqMx+NjKisrvSaTSV27dm2TqqqxDRs28G4KsixDr9fj8ccf53bzT6QnnnhCyFVVVch1u91obW3FypUrud1AIIDjx49j9erV3G4sFkNVVRXq6+u53VQqBYvFgvXr13O7uVwORqMRGzdu5HbzyxNFa8QYE3LzrS6XL1/O7Xq9XjQ3N2PVqlXcbjgcRkNDA9asWcPtJhIJvPnmmwU/nju43/zmN89985vfPMfrEQRx46CjygShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUHoLo8EoUG4g7tv376pzz777Oq2trbC72L9MdzJd3kkiJsJ97XK27dvr3U6nezVV1+9d+7cuUccDscIn89ntNvt3IMrigKv1wsRlzEGn88n5AIQdn0+H4aGhoTccDgMj8cj5CYSCeF9lclkhF1ZljVXo6GhIeEaBYNBYTcWiwnXN5VKIZfLFfx47uA+++yzp7ds2fKg0Wj05XI5yeVyVQYCAcPAwADvpqCqKoLBIETc/J38RVwAwm4wGITP5xNyo9GosJtMJuH3+4XcbDaLQCAg5CqKMqwaiY4LiNfI5/MJ7+dwOCy8n+PxuLCbTqe5ll1yB3fEiBH6Bx54wLV8+fJzRqORLVy4sGtgYCD12c9+lndTkGUZTqcTIi5jDP39/cJub2+vkOt2u1FcXCzkBgIBMMaE3FgshkQiIeSmUin4/X4hN5fLweVyCbmqqsLhcNzyGjmdTpSXlwu5Xq8XBoNByA2Hw8hms0LuTV/WV1tb66mtrS28dTlBEDccOh1EEBqEgksQGoTO4xKEBqH+uAShQegdlyA0CL3jEoQGoYNTBKFBKLgEoUHoOy5BaJCCgpvJZPQNDQ13ezwey9tvvz2jqalp9I0YnL7jEoQYBV3yqCgKfvazn9319NNPW3bu3Dkqk8kU7dq1a5dOp2OqquoYY1AUhXvwfIMkERf4IPi32h3OnIfrDmfOt+Pvzbd6uR010mp9C6Wg4JaUlChTp051AzDPnz/f39/fPw4ALjf9um+4Tb9Em0INp+mXaEOpSCSC/v5++P1+bjff9CsajXK7w2n6lcvl0NLSMqymXyLcrhoNt+lXb28vQqEQt5tMJmGz2f73NP1Kp9NGi8Uy9fz585He3l7zmDFjXDqdjplMJkZNvwrndjb9KikpuS1NvyRJuuU1oqZflzGbzbl///d/38o9G4Igbgp0VJkgNAhdOUUQGoQuwCAIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1CwSUIDULBJQgNUlBwY7GY8Qc/+MG61tbWsX//93+/9tixYyNv9sQIgvh4CrpWubS0NDcwMOBljJlramr8Fy9etC5ZssQHAP39/SO9Xq+xp6eHe/B80y8RlzEm7AIf9JcRcX0+H9xut5AbDocxNDQk5CYSCXg8HiE33/RLxM03/RJxVVW9LTUaGhoSrlF+ZZGIG4vFhOt705p+mUwmYzqdNldUVBiz2awZAHK5nDQ0NFQeCoUMg4OD3JNVVRWhUAgiLmNM2AUg7AaDQQQCASE3Go0Ku8lkEsFgUMjNZrPCbr7pl5Zq5PP54Pf7hdxwOCxco3g8LryvbkrTr0wmY1i+fLk6ODhoraioUKuqqiwAYDQa2YIFC7odDkdq2bJl3JOVZRkDAwMQcRlj6OvrE3btdruQ63a7YTabhdxAIABVVYXcWCyGeDwu5KZSKfh8PiE3l8thcHBQyFVVFf39/be8Rk6nE2VlZUKu1+uFXq8XcsPhMDKZjJB7U5b1mUwm+Stf+co73LMhCOKmQEeVCUKDUHAJQoPQQnqC0CC0kJ4gNAh9VCYIDULBJQgNQsElCA1CwSUIDULBJQgNUnBwE4mEWZZlXTweNyeTSf6+EARB3DAKCm4kEimqr6/feOLEidof/OAH/+fZZ5996GZPjCCIj6ega5XLy8uzEydOtI0YMSJbX19v+/Wvfz0LALLZrO7w4cNzz5w5Y7FardyDq6qKM2fOoKysjNtljOHMmTOoqKjgdgHg7NmzqKys5PZCoRB6e3uRyWS43VgsBpvNJnT+OpVK4ezZszAajdxuNptFS0sLSkpKuF1FUdDS0oLS0lJuN1+j8vJybhcQr1EgEIDT6UQymeR2I5EIurq6hBqkJRIJtLe3CzU5y2QyXM3CCgpuNpvVx+Pxie+++27p73//+7nPPPPMmwBQVFSkzp8/v8flcqWWLl3KPVlFUeDxeCDiMsbgcrmEXOCDFSQirtfrhcViEXJDoRB0Op2QG4/HkU6nhdxMJoNIJCLkyrIMn88n5KqqCrfbLVyjwcFBIdflcqGyslLI9fv9KCoqEnKj0SgURRFyk8kk3n///YIfX1Bw9Xq9+tOf/vSooihsxYoVHWaz+cpLQ0VFRcJisahVVVXck5VlGRaLBSIuY+y2uJlMBqWlpcLjWq1WIddoNAq7qVRK+O/N5XLCrqqqw6pRSUmJkJtMJoX3lSzLwvXV6XTC45pMJq6WooUGl02ePNnFPRuCIG4KdDqIIDQIBZcgNAgFlyA0CAWXIDQIBZcgNAgFlyA0CAWXIDQIBZcgNAgFlyA0SEHBzWazuqamphl+v9/a1NQ0vbW1dcSNGJzu8kgQYhR0yWMul5N++MMfzv3nf/5n5eLFixX79++/Z8+ePTsAgDEmXf4v9+CMsSs/Iq7ouP9zGyLerR5Xi3/vcOr7l9sQdW5HfYf79xZKQcG1WCzKtGnTnJWVlXJtbW1AkqTJAJDJZHSvvfbagvfff79UZBmUqqpobGwULs7x48e5vb90Rd7xw+EwBgYG4PV6ud14PI6uri6EQiFuN51Oo729nWvpV55sNoszZ84gm81yu4qioKmpCaqqcrv5+oownBrlO+65XPyX10ejUfT29iIQCHC7yWQSNpsNsViM281kMlzLAQsKbjqdNhoMhtq9e/eO0ul0yl133RUAAJPJpNbX159kjMU2bNjAPVlZlmEwGPD4449zu4wxSJKEJ598UsgFIOS63W60trZi5cqV3G4gEMDx48exevVqbjcWi+HQoUOor6/ndlOpFKxWK9avX8/t5nI5FBUVYePGjdyuqqrQ6/V44oknuN3h1MjpdMJms2H58uXcrtfrRXNzM1atWsXthsNhNDQ0YM2aNdzuTWn6ZTabc7/85S9fyu9MvV5PdzIniNtIwf1xdTodhZUg/pdAvYMIQoNQ7yCC0CB0AQZBaBAKLkFoEPqOSxAahL7jEoQGoY/KBKFBKLgEoUEouAShQQoKbjweNz733HNf9Hg8VYcPH57ym9/8Zt7NnhhBEB9PQZc8mkwmuaOjI3L+/PlRv/zlL+cqipJ5+umnWwDA7XZXhEIhyel0ci89URQFwWBQcjqd3EepGGO4PK7QES5R1+v1wu/3C7nhcBiBQEDITSQSwvsqk8kIu7IsD6tGoi4gXiO32y1co0AgIOzGYjHh+qZSKZVnhV1BwTUajcxqteKdd94ZV11dPbazs1MfDAZ1paWlrLu7e/Tg4ODLNpvtDO9kVVXVuVyue20222lelzEmuVyuz9hstmZeFwBcLtd9Im4wGCwbGBgYa7PZLvK6sVjMMjAw8CmbzXae102lUman0zndZrO18rrZbNY4ODhYZ7PZzvK6iqLoXS7Xp202WwuvO8waCbt+v7/C4XCMstlsnbxuOBy2DgwMTLTZbBd43UQiUex0Ou+y2WztvG46nS7KZrNFBQt/udj5435SqZTh97///co9e/YsiEajxX/605+m5H+XzWb1b7755pxCtvM/f2RZlvbt23e3iMsYw549ez59q12fz2c9duzYXSJuOBwubmhomC7ixuPxosOHD88ScdPptOHAgQNCNcrlcrr9+/fPFd3Pe/fuveX1HRoaKmtsbJwi4gYCgZL33ntvmogbjUZNb7/99kwRN5lMGt96663ZhT6+oO+4ZrNZ/trXvvbW6tWrm0pLS1OPPfaYPf+7hoaGCT6fz5JIJLi71Dc1NX2qrKxM6CqM3bt3Tw2HwyWhUKjwV6nLvPHGGxNCoZDV7/ebed3i4uJET09PWSQS4W426/V6R7rd7nHRaNTC6/p8Pp3T6aw8duzYSF63o6Ojwu/3j3nnnXem8brvvvvup7xeb1lfXx/3fj569Ohkr9dbHIlECl6Flk6nDdu2bVvW29tbGQqFivbu3TulUDebzeq2b9++JJFImDo6OmrOnTtXXagry7K0Y8eO+4PBYLnD4ajavXv39EJdVVWl3bt3L3C73SNjsVjxzp0767LZbMHP6z/96U/znE7nmFmzZvVt27bt3kKcYR1VDgaD+k2bNt136NChshMnTozl9T0eT+y1114reAf9JVVVVb4jR45UHjx4cBSvm0wmI5s3b76roaFhDK975MiR8d/73vc+5/V6ubs1nz59eup7771nyWaz3LeT+PWvf33vvn37xvT19XG706ZNC7e1tcl2u52r+3gikZB++MMfLvzXf/3XuX19fVwdppPJpO7HP/7xgt/97nejDh06VHD4JElSX3vttZLW1ta6xsbG6q1bt96nKEpBIZAkie3du7fI4XCMttvtRX/+859HFzquwWBghw8f1nV1dY2tqqqyv/jii/OSyWRBLzg6nY4dPXpUbW9vH28wGKK//e1vF6XT6YJfrJqamnKNjY1T/vM///Pe55577n6/33/dv3dYwWWMSUVFRWzkyJFJRVG4W6VPmjQpJkmSIjK2w+GwJBIJ0+c//3nue8jU1dWl586dOxiLxUy87tGjRz8zbty4MofDUfCTIs/o0aPPGI3G6H//93/P4XXj8Xjp2LFjM83NzZ/hdQ0Gg9LR0THys5/97CUeT1EUyWg0ZkeMGOHOZrNcnzBKSkrUb33rW2dNJlNNLpcr+ElsMpnUsWPHRhhjhvHjx6dqamoKHtNoNLLx48dHzGazumDBgijPfAFgwoQJYb1ez9rb2++aOHFid0lJScFHiyZNmhTW6XSora0NXO4fXXC2Jk2aFGltbS1pa2u722q1jnW73df9JDis4FZXV8srVqzoKSoqqrz33nudvH40Gp1cUVFR2dfXx/3u1dvbW7d48WIpGo1yf9x95513aseMGVO8dOlSD6/7/PPP7/3Od76zc9asWf28bmdnZ1V5eXnlww8/3Mfrfvvb326prKw0LFy4sIvXDQaD+iVLlvRNmjQpyeOVlZWpf/3Xfz0wb948ddSoUT7ecUePHq1bunRp/7Jly3oKdTKZjGH8+PHj+/r6TPF43Dp58uSeQu+4ksvl9GPGjJnQ0tIyuaenZ4LVap2sqmpB79ayLOuqq6snnT9/forT6Zy9ePFioyzLBbmqqkpWq/VTPT09Ew4cODB/0aJFLqPRmCvEBQCTyfSp8ePHl27fvv3Fr3/963snTZqUvp4jMUbXCxOE1qArpwhCg1BwCUKDUHAJQoNQcAlCg1BwCUKDUHAJQoNQcAlCg1BwCUKDUHAJQoP8PzS/I+ovoNUlAAAAAElFTkSuQmCC"
      },
      "metadata": {
       "needs_background": "light"
@@ -692,14 +671,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [6, 7] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        6,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -720,10 +699,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAoS0lEQVR4nO2deZRU5Z33v7eWruqu6o1ma/YAzdqgQkRahLyaIBgUJTYmDKhJJoszepwkJ2fmJCeZzMwx7zhhJmcySeBkToyJIoqihEUQorQL0DRNA71R0EtVL9VVXfu+33uf9w8t3hhZ6nmaJRd/n3P6eLDu5z5P3ae+tdz7PPcnMcZAEIS20N3oDhAEwQ8FlyA0CAWXIDQIBZcgNAgFlyA0CAWXIDQIBZcgNAh3cM+ePVvx4osv1g0PD5dciw4RBHFlJN4JGBs2bFhZWVkpT5gwIfmjH/2oKZPJGBhj+mvUP4L41GAwGGSDwaAUtC3vzmVZNsyZM8fr9XrLM5mMbu/evZ89ffr0M7W1tdW8+1JVFR0dHVi4cCGvCsYY2tracMstt3C7ANDa2irkxuNxeDwezJgxg9tNpVIYHBzErFmzuN1MJgO73Y65c+dyu7Is4/z585g/fz63q6oqOjs7sWDBAm6XMYb29vbrPr7RaBR+vx/Tp0/ndhOJBFwuF2pqarjddDqNvr4+zJkzh9vN5XKsuLj4/65fv357QQJjjOvv1VdfnfaNb3zjkebm5jGMMWQyGeO2bdvamAC5XI698MILIipTVZX9/ve/F3aff/55IdflcrEDBw4IuX6/n+3evVvIjUaj7LXXXhNyk8kk2759u5CbzWbZiy++KOQqisL+8Ic/CLkjGaPBwUF26NAhIdfj8bB9+/YJuaFQiL3xxhtCbjwel3fs2PEEKzCH3J+469ev71u/fn0fr0cQxNWDzioThAa5ocGVJOlGNk8QmuWGBpfRkkKCEIK+KhOEBqHgEoQGoeAShAah4BKEBqHgEoQGoeAShAah67gEoUG4g3vkyJHJP/rRjzZejcbpOi5BiME9V3nq1Klum82mqqoqqaqKs2fPThoYGDCdOXOGu3FVVTEwMAARlzEm7ALA4OCgkOv3+2G324XcSCQCh8Mh5CaTSfT19Qm52WwW/f39Qq6iKMLujRojj8cjfKxCoZDw+MbjceHxTafTyGQyBW/PHVxJkoyZTKYsEomYS0tL06qqSowxqKrKuyuoqnrhTwTRdv+8bZE2RdsdiauqqubckTzfvP9pOs48cAdXp9OZv/Wtb3UFAoGSysrKVG1t7aDNZsssWrSId1eQZRk2mw0iLvtoraeo29bWJuS63W4oiiLkBgIBRKNRITcWi2F4eFjITaVScDgcQm4ul8P58+eF3Pxa3us9Rk6nE3q9Xsj1er1Ip9NCbjgcRiAQEHITiQScTmfB23MHd8KECaG1a9c28HoEQVw96HIQQWgQCi5BaBAKLkFoEAouQWgQmjlFEBqE7oBBEBqEvioThAah4BKEBqHgEoQG4Q5uKBQyeb3eclmW6cwSQdwguKc8/va3v6222Wy3r1ix4sxXv/rV7mvRKYIgLg93cB977LHBp556an5ZWVkyX/Tr+PHjVr2ev2Cfqqo4fvw4jEYjt8sYQ1NTE0wmE7cLAE1NTTCbzdxeOBxGf38/wuEwtxuPx3H+/Hkkk0luN5VKoa2tDbIsc7vZbBYnT54UuvymKAoaGxthMHC/VC6Mb1FREbebH1+RMQoEAnC5XAgEAtxuNBpFb28vYrEYt5tMJtHZ2cm1PC9PJpPhGlvu0fjpT3+6pLi4eNq8efMaTSaTeu+997bG4/HEgw8+yLsrKIqCVCoFEZcxhng8LuQCH662EXGHh4fR3t6OlStXcrvBYBAVFRVYs2YNtxuLxWAymYT6nE6nAUDIlWUZmUxGyFVVFclk8rqP79DQEM6fP4977rmH2/X5fGhpacHq1au53UgkAovFgrVr13K7yWQSf/rTnwrenju4//qv/3omm82et1gsMQAwm805o9HIiouLeXcFWZZRVFQEEZcxNiLXaDQKuWazWbjdkbgjOVYAhJ9vLpcTbldV1U/VGGUymREdK52u8FNO3MGtrKxMAUjxegRBXD3ochBBaBAKLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQ7uAePXp09L59+xb7/X7+2d8EQVwVuKc82u32zHPPPTfrkUceSf/93/99pyzLelVVhVatKIoCRVGEXMaYsAt8ODf0evd5JK4syyNqV/T5jqRdVVVHNL4j6bMWx5fnHmzcwV26dGmqubnZm0wmjR8t61vc2NhYyrsf4P8v+xJZEgh8uDRPZEkgAOHlhJFIBAMDAwgGg9xuIpFAV1cX4vE4t5tOp9He3o5sNsvt5nI5tLS0CBejOn78ONcE+DyMMRw/flxoSSAgPr6hUAgulwter5fbjcVisNvtiEQi3G4qlcLZs2eRSvFP5ecdV+4junfv3s/o9frKuro6m8lkUtetW3dCluXYxo38JXNlWYZOp8Ojjz7K7ebfkUVdRVGEXLfbjdbWVqFlX4FAAEePHhVa9hWLxVBZWYn6+npuN5VKobi4GBs2bOB2c7kc9Ho9Nm3axO3m3yiu9xg5nU7YbDahpZderxfNzc1CSy/D4TAaGhqwbt06bjeRSODNN98seHvu4H7ve9/rBnBV7nxB91UmCDHovsoEoUHochBBaBAKLkFoEKodRBAahH7jEoQGoa/KBKFBKLgEoUEouAShQSi4BKFB6KwyQWgQ7uBGIhHdd77znQecTmfVSBuns8oEIQb3XOXdu3fPfPnll6d84xvfaK2urg729vaOc7vdxnPnznE3rqoq3G43RFzGmLALfFgDSMT1+/1wOp1CbiQSwdDQkJCbSCTgcrmE3EwmI3ysFEXR3Bh5PB7hMQoGg8JuLBYTHt9UKsW1Qog7uOFweNLMmTPHOByO6tmzZw+Gw2FLIpHQiyxzU1UV8XhcaIkc8OGLWdQVbTccDiMWiwm50WhU2E2lUsJ9zmazwq6iKMIuY2xEYyTqjmSMRuImEgnhY5VOp7mWXXIH9+mnnz58++23d86ZMyduNBrZ4sWL7Xa7PX3nnXfy7gqyLMPhcEDEZYyhu7tb2O3q6hJy3W43DAaDkBsIBJDNZoXcWCyGUCgk5KZSKbjdbiE3l8uhv79fyFVVFb29vdd9jJxOJ0pKSoRcr9cLxpiQGw6HEY/HhdxEIgG/31/w9kIrnOvq6jwiHkEQVwe6HEQQGoQuBxGEBqFFBgShQeirMkFoEAouQWgQCi5BaBAKLkFoEDqrTBAahM4qE4QG4Q5uf39/9YEDB5bG4/Hia9EhgiCuDPeUx/fff39+c3Ozua6ujgHAR0W/pFwux914vkiSiJsvUSHi5tsWcfNFsMi9MvmiX6JjpKqqpp7vSN1rWvRr0qRJp48cObLoueeem/fUU0+d+ajol1Xk92q+oJRIUSjGGJqamlBUVMTtAsCJEydgMpm4vXA4jIGBAYRCIW43Ho+jq6sLiUSC202n02hraxN6UeRyOZw8eZLbAz58gxMtzJYfI9HCbKLjGwwG4XK5uCbt54lGo7Db7YhGo9xuMpnE2bNnkU6nud1MJsO1PXdizpw5U2GxWMrr6uo6r0bRL0mShApK5Yt+ibqyLAu5N7LoV0VFhXDRL7PZfEOKfjHGrvsYUdGvi/Dd7363F0Avr0cQxNWDruMShAah4BKEBqHgEoQGoeAShAah4N7E5M/MptNp5HI55HI5mq12kyB0zynirxvGGCKRCE6ePAmPxwNJkpBKpfDqq6+irKwMixcvRnV1Nc0V1zAU3JsMxhjOnDmDjo4OLFmyBCtWrLgwwUVVVQwPD+Po0aOoqqr62GOEtqCvyjcR+dAODAzgkUcewaxZs1BUVASdTgedTgeDwYBJkybhS1/6EkwmE95++22ue/kSfz1QcG8iQqEQOjo6cN9998FkMl3yq7Ber0ddXR1kWUZPTw/97tUgFNybiJMnT6Kurq6gucE6nQ4rVqxAS0sLfepqEO7gvvvuu2P+6Z/+ae2uXbvGXosOEWIoigKPx4OpU6cWfNLJarVCp9MhlUpd494RVxvuMxP/8z//s1SSpOLbbrtNUhRF6unpGe9yuYpsNht344qiwOVyQcRljAm7wIeLBURcn8+HgYEBITccDmNwcFDITSQSFybPX4xUKgVFUbhONkmSBIvFgpaWFowde/H3YVmWhY+zqqojGl/RMRoeHhYeo2AwKDxGsVjssmN0Oa550a8pU6ZkAKQ7Ojrm19fXe2OxWHEqldJFIhHeXUFVVSSTSYi4jDGkUikhF4CwG4vFhPs8EjeZTF62z5lMBrIsc+9XURQkEolL7ldRFBqjAonH48Lu9Sj61Xjy5Mn5NTU1NoPBwG677TZHT09PeunSpby7gizL6O3thYibLwol6p47d07Idbvd0Ov1Qm4gEEA6nRZy8xXkLuWqqorBwUFks9mC1xnnq+k99NBDKCkpueg2uVwODodDqM+qqqK7u/u6j5HT6URxcbGQ6/V6oSiKkJuv9CfiJhIJeL3egrfnDu706dNj06dPP87rEdcWSZIwZcoUdHV1oba2tqDfucFgEEajEWaz+Tr0kLia0FnlmwRJkrBo0SKcOnUKyWTyitvLsoyGhgYsXboUOh29DLQGjdhNhMViwZ133ok9e/YgGo1e9PosYwyZTAaHDh3ChAkTMGnSpBvQU2Kk0Hy3mwhJkjBz5kwYDAbs2rULNTU1mDt3LsrKyiBJEhKJBBwOB86cOYPa2lrceuutNF9Zo1BwbzIkScK0adNQXV2Nc+fO4fDhw4hGoxgaGkJNTQ0mT56MdevWwWq1Umg1zA0N7qfhhcMYQy6Xg8vlwvnz53Hu3DnMnj0bkydPRnFx8WWPQd71er3o7+9Hf38/7HY7qqurYTabL+lKkgSTyYR58+Zh9OjRGBwchMfjwe23335F9y/77HQ60d3djfPnz3P32ePxYHBwsKA+/2W7LpcLPT09OH/+PKZMmXJF99PGDQ3uzTxHNn/f53fffRdbtmxBU1MTfD4fFEXB5s2bMX/+fHz9619HfX39J8KQd0+ePImuri6MHTsW48aNw+c//3n09fXh2LFjGDNmDJYvX35RV5ZltLS0fMxdvXr1BXfs2LG46667LhrCXC53oc8nTpy40OctW7agtrYWf/u3f4svfelLBbe7Zs2aK7abdxsaGrB169aPtfub3/zmsu1+amGMjegvk8kYt23b1sYEyOVy7IUXXhBRmaqq7Pe//72w+/zzzwu5LpeLHThw4IrbJZNJ9s///M+svLycAbjon8lkYps2bWLDw8OfcF999VV25MgRlk6nmaqqH+t7LpdjnZ2d7A9/+APz+Xwfe7wQt6Oj45LuT37yE1ZWVnbZPm/cuJF5PJ5PuDt27CioXb/ff+FxVVVZIpFgP/7xj6/Y7qOPPsq8Xu/H9n0xBgcH2aFDh644RhfD4/Gwffv2CbmhUIi98cYbQm48Hpd37NjxBCswd/Qb9xqQ/6R49tlnLzuNLZPJYPv27TAYDNiyZQuKi4uhKAoOHTqEefPmYd68eZ/4dJEkCQaDAXPnzsWoUaNw4MABrF+/HmazGYqi4ODBg1d0582bh6qqqk+4W7duxbPPPnvZm3NnMhm8/PLLKCoqwpYtW2A2m6GqKt566y3U1tZi7ty5l2131KhR2L9//4V2GWPYunUrfvazn12x3Zdeegkmkwm//OUvP/XXnuly0FWGMYbz589j8+bNBc09VVUVO3bswP79+8EYg8PhuPAiv9xXQkmSMG7cOCxYsADHjx8HYwx2ux1FRUVcblNT08f6XMgd9VVVxSuvvIK33noLjDH09PTAbDZjzpw5V2x3/PjxmD9/Pk6cOAHGGGw2G1e727Ztw6FDh27qn1mFwB3cWCxmDIVCpalUSqyuxKeAXbt2we12F7x9KpXCCy+8AEVR0NraiiVLlhT0O06SJMybNw8DAwOQZVnYzeVy2LlzJzweD1efX3zxRWSzWbS2tuKOO+4oaCKHJEmora1Ff38/stksXn/9da6pful0Gtu2bROal30zwR3ctra2cU888cSXX3rppdpr0SGto6oqjh49yv2JcPbsWXi9XiSTSVRVVRXsGY1GlJWVwe/3I51OY9SoUVyu1WpFIBDAyZMnuft8+vRpBINB5HI5lJeXc7VbWlqKUCiEI0eOcLd76tQp4YULNwvcv3GXLVvm/NnPfhZevnx5byaT0R04cOC2EydOWEUKaKmqiubmZhQX81fsZIyhubkZFouF2wU+XHRutVq5vXA4jL6+PsTj8Ys+nsvl0NfXx71fj8eDV155BSaTiWsKoiRJMBqNeO2112A0GrnOuEqSBL1ej9dee014meLOnTsve7eNy7X9xhtvoKuri7tdj8eDN95445JvUoFAAENDQ0Lhjkaj6OnpEVqjnEwm0dHRAUVRuN1MJsNV+Is7uF1dXVaz2SzPnj07CgArVqywBQKB5KpVq3h3BUVREIlEIOIyxhAMBoVcAPD7/UKux+NBe3s7vvCFL1z0cUVR8Nxzz3EHoaqqCmvWrEFjYyNUVS04vOyja5+rV6/G8eN8az/YR5ed7rvvPuzfvx89PT1cfmlpKVatWnXh9ypv2/feey927NiB/v5+LreqqgorV67E6NGjL/r40NAQurq6cPfdd3PtF/jwdVFWVib02ohEIjAYDEJuMplEQ0NDwdtzB3fixInJX/7yl2/l/221WtNms1ktLS3l3RVkWYbZbIaIyxi7IW48HkdxcfElXcYY6urq8M4773Dtd/bs2Zg+fTra2toQDocL/soryzKi0SimTZuGM2fOIBKJoKKiomA3Ho9jypQpWLx4MQ4ePMjV5wULFmDy5Mk4ffo04vF4wV+XFUVBPB7HxIkTcdddd+GDDz7gCv6tt96KiRMnXrIEp9VqvewYXY5UKiXsKooi7OZv6Ffw9rwNWCwWdezYsfwFQC/CzXoh/eGHH8aYMWMK3r6oqAibNm2CwWBAbW0tWlpaCn4h9/T0oLq6GkajEbW1tWhubi7IZYyhu7sb1dXVKCoqQn19Pddv66KiImzcuBFms5m73XPnzl2YSfXwww+jsrKSq90NGzYI19y9Wbihl4NuxlP6+bOmTz31VEG3kZEkCQ888AAefPBBSJKEmpoahMNh2O32yx6f/E+FEydO4M4774QkSZg1a1bBbigUQnNzM+rq6iBJEhYsWIAnn3yyoEBIkoS1a9figQcegCRJmD17NoLBIBwOR0HttrS0YOnSpZAkCQsXLuRq96GHHsKaNWtu2jf9Qrmhwb1ZD77BYMB3v/tdPPHEE5c98abX6/HFL34RP//5zy98vTIYDLjvvvtw/PhxnDlzBrIsfywM7KOC3v39/dizZw9Wrlx54e4Vf+kqinJRt6+v76Lu97//ffzd3/3dZSc35Nv4+c9/fuHEYP7/HTt2DK2trZds1+FwYM+ePbj33nsvtKvX6/H9738f3/72t6/Y7v3334//+q//Ej4heTNBc5WvEVarFZs3b8by5cuxdetWtLa2IhKJgDEGq9WK6dOn4/HHH8fXvva1T/w2tFgsqK+vx9GjR/HKK69gypQpqK6uhl6vh8/nQ19fH0wmE9atW3dhyd7F3JdffpnLtVqt+I//+A8sW7YMv/71r9He3v6xPs+YMQNf/epX8fjjj6O8vPwT7uXadTgcKC4u/kS7kiShtLQUmzdvxrJly7B161audj+t0JTHa4QkSTCbzVi/fj3WrFkDu92Ojo4OdHZ2Yu3ataipqUF5eflFT0jkV/fcfffdSCaTcDqdcDgcaGlpwf33349Vq1ahrKysYHdgYACNjY148MEHr+jm+/zFL34Rdrsddrsd77//Pr7yla9csc9msxn33HPPhXZdLhcaGhrw8MMPY/Xq1Vds98tf/jLuv/9+2O129PX14Z133sHGjRsv2+6nFQruNSZ/C9QFCxZgwoQJsFgsWLJkCZc7e/ZsTJgwAdFoFAsWLOB2p0yZAq/Xy+VarVYsXLgQc+fORTweF+pzTU0N+vv7hdpdsGABgsFgwe1+2qC3MILQIBRcgtAgFFyC0CAUXILQINzBHRgYKHv77bdv6erquvit7zmg0/oEIQb3WeWf/OQnt7/33nuznnjiCd0//uM/nlZVVcpfYOdFVVWIuvlbeIiWiLwR7Y7EvVHHaiTtarHPIx1fVVVH1OdC4Q6uoiisqqoqYDAYTJlMRrdr164lR48eLRXtbGNjI7cHfHiQjh07Jnxt79ixY9Dr9dxeOBzGwMAAfD4ftxuPx9Hd3S1cFKq9vV1ouVk2m8WpU6eEi4Ll75LBi6qqOHbsmNA3K8YYGhsbhcYoGAzC7XZjeHiY241Go7Db7QgGg9xuMpnE2bNnkUgkuN1MJsP1WuYO7vLly4PxeHx8ZWWl12QyqQ8//HCTqqqxjRs38u4KsixDr9fj0Ucf5XbzL6THHntMyFVVVch1u91obW3F6tWrud1AIICjR49i7dq13G4sFsOoUaNQX1/P7aZSKVgsFmzYsIHbzeVyMBqN2LRpE7ebX54oOkaMMSE3X+py5cqV3K7X60VzczPWrFnD7YbDYTQ0NGDdunXcbiKRwJtvvlnw9tzB/eY3v3nmm9/85hlejyCIqwedVSYIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1Cd3kkCA3CHdy9e/fOePrpp9e2tbUVfhfrS3Az3+WRIK4l3HOVt2/fXuN0Otmrr766aOHChe8MDAyM9vl8Rrvdzt24oijwer0QcRlj8Pl8Qi4AYdfn82F4eFjIDYfD8Hg8Qm4ikRA+VplMRtiVZVlzYzQ8PCw8RsFgUNiNxWLC45tKpZDL5Qrenju4Tz/99MmtW7feYzQafblcTnK5XJWBQMAwODjIuyuoqopgMAgRN38nfxEXgLAbDAbh8/mE3Gg0Kuwmk0n4/X4hN5vNIhAICLmKooxojETbBcTHyOfzCR/ncDgsfJzj8biwm06nuZZdcgd39OjR+rvvvtu1cuXKM0ajkS1durR7cHAw9bnPfY53V5BlGU6nEyIuYwz9/f3CrsPhEHLdbjeKi4uF3EAgAMaYkBuLxZBIJITcVCoFv98v5OZyObhcLiFXVVUMDAxc9zFyOp0oLy8Xcr1eLwwGg5AbDoeRzWaF3Gu+rK+mpsZTU1NTeOlygiCuOnQ5iCA0CAWXIDQIXcclCA1C9XEJQoPQJy5BaBD6xCUIDUInpwhCg1BwCUKD0G9cgtAgBQU3k8noGxoabvF4PJa33357TlNT07ir0Tj9xiUIMQqa8qgoCv793/995pNPPmnZsWPH2EwmU/Taa6+9ptPpmKqqOsYYFEXhbjxfIEnEBT4M/vV2R9Lnkboj6fONeL75Ui83Yoy0Or6FUlBwS0pKlBkzZrgBmJcsWeLv7++fCAAfFf26faRFv0SLQo2k6JdoQalIJIL+/n74/X5uN1/0KxqNcrsjKfqVy+XQ0tIyoqJfItyoMRpp0S+Hw4FQKMTtJpNJ2Gy2v56iX+l02mixWGZ0dHREHA6Hefz48S6dTsdMJhOjol+FcyOLfpWUlNyQol+SJF33MaKiXx9hNptz//mf//kid28Igrgm0FllgtAgNHOKIDQITcAgCA1CwSUIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1CwSUIDVJQcGOxmPEHP/jB+tbW1gn/8A//8PCRI0fGXOuOEQRxaQqaq1xaWpobHBz0MsbM1dXV/nPnzlnvuusuHwD09/eP8Xq9xt7eXu7G80W/RFzGmLALfFhfRsT1+Xxwu91CbjgcxvDwsJCbSCTg8XiE3HzRLxE3X/RLxFVV9YaM0fDwsPAY5VcWibixWEx4fK9Z0S+TyWRMp9PmiooKYzabNQNALpeThoeHy0OhkGFoaIi7s6qqIhQKQcRljAm7AITdYDCIQCAg5EajUWE3mUwiGAwKudlsVtjNF/3S0hj5fD74/X4hNxwOC49RPB4XPlbXpOhXJpMxrFy5Uh0aGrJWVFSoo0aNsgCA0Whkd9xxR8/AwEBqxYoV3J2VZRmDg4MQcRlj6OvrE3btdruQ63a7YTabhdxAIABVVYXcWCyGeDwu5KZSKfh8PiE3l8thaGhIyFVVFf39/dd9jJxOJ8rKyoRcr9cLvV4v5IbDYWQyGSH3mizrM5lM8le+8pXD3L0hCOKaQGeVCUKDUHAJQoPQQnqC0CC0kJ4gNAh9VSYIDULBJQgNQsElCA1CwSUIDULBJQgNUnBwE4mEWZZlXTweNyeTSf66EARBXDUKCm4kEimqr6/fdOzYsZof/OAH/+fpp5++91p3jCCIS1PQXOXy8vLslClTbKNHj87W19fbfvWrX80DgGw2qzt06NDCU6dOWaxWK3fjqqri1KlTKCsr43YZYzh16hQqKiq4XQA4ffo0Kisrub1QKASHw4FMJsPtxmIx2Gw2oevXqVQKp0+fhtFo5Haz2SxaWlpQUlLC7SqKgpaWFpSWlnK7+TEqLy/ndgHxMQoEAnA6nUgmk9xuJBJBd3e3UIG0RCKB9vZ2oSJnmUyGq1hYQcHNZrP6eDw+5d133y393e9+t/Cpp556EwCKiorUJUuW9LpcrtTy5cu5O6soCjweD0RcxhhcLpeQC3y4gkTE9Xq9sFgsQm4oFIJOpxNy4/E40um0kJvJZBCJRIRcWZbh8/mEXFVV4Xa7hcdoaGhIyHW5XKisrBRy/X4/ioqKhNxoNApFUYTcZDKJDz74oODtCwquXq9Xf/rTn76nKApbtWpVp9lsvvDWUFFRkbBYLOqoUaO4OyvLMiwWC0RcxtgNcTOZDEpLS4XbtVqtQq7RaBR2U6mU8PPN5XLCrqqqIxqjkpISITeZTAofK1mWhcdXp9MJt2symbhKihYaXDZt2jQXd28Igrgm0OUggtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIAUFN5vN6pqamub4/X5rU1PT7NbW1tFXo3G6yyNBiFHQlMdcLif98Ic/XPgv//Ivyrlz5yr27dt32+7du18BAMaY9NF/uRtnjF34E3FF2/3LfYh417tdLT7fkYzvn+9D1LkR4zvS51soBQXXYrEos2bNclZWVso1NTUBSZKmAUAmk9G9/vrrd3zwwQelIsugVFVFY2Oj8OAcPXqU2/tzV+QTPxwOY3BwEF6vl9uNx+Po7u5GKBTidtPpNNrb27mWfuXJZrM4deoUstkst6soCpqamqCqKrebH18RRjJG+Yp7Lhf/9PpoNAqHw4FAIMDtJpNJ2Gw2xGIxbjeTyXAtBywouOl02mgwGGr27NkzVqfTKTNnzgwAgMlkUuvr648zxmIbN27k7qwsyzAYDHj00Ue5XcYYJEnC448/LuQCEHLdbjdaW1uxevVqbjcQCODo0aNYu3YttxuLxXDw4EHU19dzu6lUClarFRs2bOB2c7kcioqKsGnTJm5XVVXo9Xo89thj3O5IxsjpdMJms2HlypXcrtfrRXNzM9asWcPthsNhNDQ0YN26ddzuNSn6ZTabc7/4xS9eyB9MvV5PdzIniBtIwfVxdTodhZUg/kqg2kEEoUGodhBBaBCagEEQGoSCSxAahH7jEoQGod+4BKFB6KsyQWgQCi5BaBAKLkFokIKCG4/Hjc8888xDHo9n1KFDh6b/+te/XnytO0YQxKUpaMqjyWSSOzs7Ix0dHWN/8YtfLFQUJfPkk0+2AIDb7a4IhUKS0+nkXnqiKAqCwaDkdDq5z1IxxvBRu0JnuERdr9cLv98v5IbDYQQCASE3kUgIH6tMJiPsyrI8ojESdQHxMXK73cJjFAgEhN1YLCY8vqlUSuVZYVdQcI1GI7NarTh8+PDEqqqqCV1dXfpgMKgrLS1lPT0944aGhl6y2WyneDurqqrO5XItstlsJ3ldxpjkcrk+a7PZmnldAHC5XLeLuMFgsGxwcHCCzWY7x+vGYjHL4ODgZ2w2Wwevm0qlzE6nc7bNZmvldbPZrHFoaKjWZrOd5nUVRdG7XK5bbTZbC687wjESdv1+f8XAwMBYm83WxeuGw2Hr4ODgFJvNdpbXTSQSxU6nc6bNZmvnddPpdFE2my0qWPjzxc6X+kulUobf/e53q3fv3n1HNBot/uMf/zg9/1g2m9W/+eabCwrZz1/+ybIs7d279xYRlzGG3bt333q9XZ/PZz1y5MhMETccDhc3NDTMFnHj8XjRoUOH5om46XTasH//fqExyuVyun379i0UPc579uy57uM7PDxc1tjYOF3EDQQCJe+///4sETcajZrefvvtuSJuMpk0vvXWW/ML3b6g37hms1n+2te+9tbatWubSktLUw8++KA9/1hDQ8Nkn89nSSQS3FXqm5qaPlNWViY0C2Pnzp0zwuFwSSgUKvxd6iN27do1ORQKWf1+v5nXLS4uTvT29pZFIhHuYrNer3eM2+2eGI1GLbyuz+fTOZ3OyiNHjozhdTs7Oyv8fv/4w4cPz+J133333c94vd6yvr4+7uP83nvvTfN6vcWRSKTgVWjpdNqwbdu2FQ6HozIUChXt2bNneqFuNpvVbd++/a5EImHq7OysPnPmTFWhrizL0iuvvHJnMBgsHxgYGLVz587Zhbqqqko7d+68w+12j4nFYsU7duyozWazBb+u//jHPy52Op3j582b17dt27ZFhTgjOqscDAb1mzdvvv3gwYNlx44dm8Drezye2Ouvv17wAfpzRo0a5XvnnXcqDxw4MJbXTSaTkS1btsxsaGgYz+u+8847k773ve993uv1cldrPnny5Iz333/fks1muW8n8atf/WrR3r17x/f19XG7s2bNCre1tcl2u52r+ngikZB++MMfLv23f/u3hX19fVwVppPJpO7HP/7xHb/97W/HHjx4sODwSZKkvv766yWtra21jY2NVS+++OLtiqIUFAJJktiePXuKBgYGxtnt9qI//elP4wpt12AwsEOHDum6u7snjBo1yv78888vTiaTBb3h6HQ69t5776nt7e2TDAZD9De/+U1dOp0u+M2qqakp19jYOP2///u/Fz3zzDN3+v3+Kz7fEQWXMSYVFRWxMWPGJBVF4S6VPnXq1JgkSYpI2wMDA5ZEImH6whe+wH0Pmdra2vTChQuHYrGYidd97733Pjtx4sSygYGBgl8UecaNG3fKaDRG//d//3cBrxuPx0snTJiQaW5u/iyvazAYlM7OzjGf+9znzvN4iqJIRqMxO3r0aHc2m+X6hlFSUqJ+61vfOm0ymapzuVzBL2KTyaROmDAhwhgzTJo0KVVdXV1wm0ajkU2aNCliNpvVO+64I8rTXwCYPHlyWK/Xs/b29plTpkzpKSkpKfhs0dSpU8M6nQ41NTWBj+pHF5ytqVOnRlpbW0va2tpusVqtE9xu9xW/CY4ouFVVVfKqVat6i4qKKhctWuTk9aPR6LSKiorKvr4+7k8vh8NRu2zZMikajXJ/3T18+HDN+PHji5cvX+7hdZ999tk93/nOd3bMmzevn9ft6uoaVV5eXnnffff18brf/va3WyorKw1Lly7t5nWDwaD+rrvu6ps6dWqSxysrK1P/5m/+ZnDx4sXq2LFjfbztjhs3Trd8+fL+FStW9BbqZDIZw6RJkyb19fWZ4vG4ddq0ab2F3nEll8vpx48fP7mlpWVab2/vZKvVOk1V1YI+rWVZ1lVVVU3t6OiY7nQ65y9btswoy3JBrqqqktVq/Uxvb+/k/fv3L6mrq3MZjcZcIS4AmEymz0yaNKl0+/btz3/961/fM3Xq1PSVHIkxmi9MEFqDZk4RhAah4BKEBqHgEoQGoeAShAah4BKEBqHgEoQGoeAShAah4BKEBqHgEoQG+X/Dcc4OaIWRHwAAAABJRU5ErkJggg=="
      },
      "metadata": {
       "needs_background": "light"
@@ -734,14 +711,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"Alice is trying to build a vertical line now. I need to block her next move otherwise she will have a chance to win. I will place my piece at [5, 7].\",\n",
       "    \"move\": [\n",
       "        5,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -762,10 +739,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAp0klEQVR4nO2deXQU15m3f9WLuiW1NsQmFkEwAgOC2BBjZAz5nASDg4NNEEkYwA6ZLJ6xj7Mcz8xxjjMZz0m+eMyMz2TiwMkkthMbY2NjCIvByItsDAghBNqgkdCuVrd63/equt8fdvPFMUvfKwQp/D7n6Phg1VP3qm7/uqur7q1XYoyBIAhtobveHSAIgh8KLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQCi5BaBDu4J49e7b4pZdeqhoaGsobiQ4RBHFlJN4JGOvWrVtWUlIiT5gwIfbEE0/UJ5NJA2NMP0L9I4jPDAaDQTYYDEpW2/LuXJZlw8033+xyuVxFyWRSt2/fvi+cPn36F5WVlWW8+1JVFW1tbZg3bx6vCsYYWlpa8PnPf57bBYDm5mYhNxKJwOl04qabbuJ24/E4BgYGMGPGDG43mUyiu7sbs2bN4nZlWUZ7ezvmzJnD7aqqijNnzmDu3LncLmMMra2t13x8Q6EQPB4Ppk2bxu1Go1HY7XZUVFRwu4lEAr29vbj55pu53XQ6zXJzc//v2rVrt2clMMa4fl577bWp3/3ud7/R0NAwhjGGZDJp3LZtWwsTIJ1OsxdffFFEZaqqsj/+8Y/C7gsvvCDk2u12dvDgQSHX4/GwPXv2CLmhUIi9/vrrQm4sFmPbt28XclOpFHvppZeEXEVR2J/+9CchdzhjNDAwwGpqaoRcp9PJ9u/fL+T6/X62a9cuITcSicg7dux4iGWZQ+5P3LVr1/auXbu2l9cjCOLqQVeVCUKDXNfgSpJ0PZsnCM1yXYPLaEkhQQhBp8oEoUEouAShQSi4BKFBKLgEoUEouAShQSi4BKFB6D4uQWgQ7uAeOXJk8hNPPLH+ajRO93EJQgzuucpTpkxxWK1WVVVVSVVVnD17dlJ/f7+pqamJu3FVVdHf3w8RlzEm7ALAwMCAkOvxeNDd3S3kBoNB9PT0CLmxWAy9vb1CbiqVQl9fn5CrKIqwe73GyOl0Ch8rv98vPL6RSER4fBOJBJLJZNbbcwdXkiRjMpksDAaD5oKCgoSqqhJjDKqq8u4Kqqpe+BFBtN2/bFukTdF2h+Oqqqo5dzh/b8b/LB1nHriDq9PpzN///vc7vF5vXklJSbyysnLAarUm58+fz7sryLIMq9UKEZd9vNZT1G1paRFyHQ4HFEURcr1eL0KhkJAbDocxNDQk5MbjcfT09Ai56XQa7e3tQm5mLe+1HiObzQa9Xi/kulwuJBIJITcQCMDr9Qq50WgUNpst6+25gzthwgT/qlWrank9giCuHnQ7iCA0CAWXIDQIBZcgNAgFlyA0CM2cIggNQk/AIAgNQqfKBKFBKLgEoUEouAShQbiD6/f7TS6Xq0iWZbqyRBDXCe4pj3/4wx/KrFbrbUuXLm369re/fX4kOkUQxOXhDu4DDzww8Mgjj8wpLCyMZYp+HT9+3KLX8xfsU1UVx48fh9Fo5HYZY6ivr4fJZOJ2AaC+vh5ms5nbCwQC6OvrQyAQ4HYjkQja29sRi8W43Xg8jpaWFsiyzO2mUimcPHlS6Paboiioq6uDwcD9Urkwvjk5OdxuZnxFxsjr9cJut8Pr9XK7oVAIXV1dCIfD3G4sFsOZM2e4ludlSCaTXGPLPRq//OUvF+bm5k6dPXt2nclkUu++++7mSCQSve+++3h3BUVREI/HIeIyxhCJRIRc4KPVNiLu0NAQWltbsWzZMm7X5/OhuLgYK1eu5HbD4TBMJpNQnxOJBAAIubIsI5lMCrmqqiIWi13z8R0cHER7ezu+9KUvcbtutxuNjY1YsWIFtxsMBpGfn49Vq1Zxu7FYDG+//XbW23MH98knn2xKpVLt+fn5YQAwm81po9HIcnNzeXcFWZaRk5MDEZcxNizXaDQKuWazWbjd4bjDOVYAhP/edDot3K6qqp+pMUomk8M6Vjpd9pecuINbUlISBxDn9QiCuHrQ7SCC0CAUXILQIBRcgtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIBRcgtAg3ME9evTo6P379y/weDz8s78JgrgqcE957O7uTj733HMzvvGNbyT+8R//8Ywsy3pVVYVWrSiKAkVRhFzGmLALfDQ39Fr3eTiuLMvDalf07x1Ou6qqDmt8h9NnLY4vzzPYuIO7aNGieENDgysWixk/Xta3oK6uroB3P8D/X/YlsiQQ+GhpnsiSQADCywmDwSD6+/vh8/m43Wg0io6ODkQiEW43kUigtbUVqVSK202n02hsbBQuRnX8+HGuCfAZGGM4fvy40JJAQHx8/X4/7HY7XC4XtxsOh9Hd3Y1gMMjtxuNxnD17FvE4/1R+3nHlPqL79u37nF6vL6mqqrKaTCZ19erVJ2RZDq9fz18yV5Zl6HQ6bNy4kdvNvCOLuoqiCLkOhwPNzc1Cy768Xi+OHj0qtOwrHA6jpKQE1dXV3G48Hkdubi7WrVvH7abTaej1emzYsIHbzbxRXOsxstlssFqtQksvXS4XGhoahJZeBgIB1NbWYvXq1dxuNBrFm2++mfX23MH9yU9+ch7AVXnyBT1XmSDEoOcqE4QGodtBBKFBKLgEoUGodhBBaBD6jksQGoROlQlCg1BwCUKDUHAJQoNQcAlCg9BVZYLQINzBDQaDuh/96Edfs9lspcNtnK4qE4QY3HOV9+zZM/2VV14p/+53v9tcVlbm6+rqGudwOIznzp3jblxVVTgcDoi4jDFhF/ioBpCI6/F4YLPZhNxgMIjBwUEhNxqNwm63C7nJZFL4WCmKorkxcjqdwmPk8/mE3XA4LDy+8Xica4UQd3ADgcCk6dOnj+np6SmbOXPmQCAQyI9Go3qRZW6qqiISiQgtkQM+ejGLuqLtBgIBhMNhITcUCgm78XhcuM+pVErYVRRF2GWMDWuMRN3hjNFw3Gg0KnysEokE17JL7uA++uij7912221nbr755ojRaGQLFizo7u7uTtxxxx28u4Isy+jp6YGIyxjD+fPnhd2Ojg4h1+FwwGAwCLlerxepVErIDYfD8Pv9Qm48HofD4RBy0+k0+vr6hFxVVdHV1XXNx8hmsyEvL0/IdblcYIwJuYFAAJFIRMiNRqPweDxZby+0wrmqqsop4hEEcXWg20EEoUHodhBBaBBaZEAQGoROlQlCg4g9fo/4m0dVVbjdbrS0tKClpQVNTU3Iz8/H3LlzUV5eDp1OR19VNAwF9waDMYZgMIjf//73+P3vf4+enh4oigLGGF555RWMGzcOa9aswT/90z9h0qRJFF6NQsG9gWCMwe1246GHHsK+ffs+9WBuRVFgt9vx7LPPoq6uDs8//zwqKyspvBqErirfQKRSKTz55JPYs2fPZZ+mzxjDyZMn8cMf/hB+v/8a9pC4WtBV5RuIhoYGvPjii1lPnTt8+DB27NhB46BBuIPb19dXdvDgwUWRSCR3JDpEiMEYw86dO7nKmyiKgh07diCZTI5gz4iRgPs77uHDh+c0NDSYq6qqGAB8XPRLSqfT3I1niiSJuJkSFSJupm0RN1ME62/NTSaTsFqt3Pvt7e2Fx+PBuHHjhNq9HJmiX6JjpKrq39xxHkl3RIt+TZo06fSRI0fmP/fcc7MfeeSRpo+LfllEvq9mCkqJFIVijKG+vh45OTncLgCcOHECJpOJ2wsEAujv7xf6bhiJRNDR0YFoNMrtJhIJtLS0XDa4fX193PsNh8PYuXMnRo8efdHfK4oiXJgtM0aihdlEx9fn88Fut3NN2s8QCoXQ3d2NUCjE7cZiMZw9exaJRILb5T3r4U5MU1NTcX5+flFVVdWZq1H0S5IkoYJSmaJfoq4sy0Lu9Sz6VVxcfMmiX+l0GgcPHkR7ezvXfkePHo0HH3wQJSUll9zvcIp+Mcau+RhR0a+L8OMf/7gLQBevR4wsmaWGu3bt4jrlWrhwISwWywj2jBgJaMrjDYIkSVi9ejWmTZuWtVNYWIhNmzYJ168lrh8U3BuIKVOm4Mknn0RhYeEVt83JycHDDz+MJUuW0P10DULBvYHQ6XT45je/iWeeeQYTJ0685HZFRUV47LHH8PjjjwtfOCKuL3SOdIOh1+uxadMmLFq0CM8//zxqamrg8XiQSCQwevRoLF68GJs2bcIdd9xBodUwFNwbDEmSIEkS5syZg6effhpPPPEEBgYG8Pbbb2Pjxo0oKSmBwWCg02ONQ8G9AcmsEDp58iScTickSYLFYsHbb7+NwsJCLFiwAGVlZRReDUPBvcFgjKGpqQltbW1YuHAhli5deuGqsaqqGBoawtGjR1FaWvqJ3xHagi5O3UBkQtvf349vfOMbmDFjBnJycqDT6aDT6WAwGDBp0iR8/etfh8lkwjvvvMP1LF/ibwcK7g2E3+9HW1sb7rnnHphMpkueCuv1elRVVUGWZXR2dtLqIA1Cwb2BOHnyJKqqqrK6WqzT6bB06VI0NjbSp64G4Q7u+++/P+Zf/uVfVu3evXvsSHSIEENRFDidTkyZMiXri04WiwU6nQ7xeHyEe0dcbbivTPzP//zPIkmScm+99VZJURSps7NzvN1uzxFZUpZ5lIqIyxgTdoGPFguIuG63G/39/UJuIBDAwMCAkBuNRi9Mnr8Y8XgciqJwXWySJAn5+flobGzE2LEXfx+WZVn4OKuqOqzxFR2joaEh4THy+XzCYxQOhy87RpdjxIt+lZeXJwEk2tra5lRXV7vC4XBuPB7XBYNB3l1BVVXEYjGIuIwxxONxIReAsBsOh4X7PBw3Fotdts/JZPKyj6u5FIqiIBqNXnK/iqLQGGVJJBIRdq9F0a+6kydPzqmoqLAaDAZ266239nR2diYWLVrEuyvIsoyuri6IuJmiUKLuuXPnhFyHwwG9Xi/ker1eJBIJITdTQe5SrqqqGBgYQCqVynqdcaaa3v3334+8vLyLbpNOp9HT0yPUZ1VVcf78+Ws+RjabDbm5uUKuy+WCoihCbqbSn4gbjUbhcrmy3p47uNOmTQtPmzbtOK9HjCySJKG8vBwdHR1ZP7nR5/PBaDTCbDZfgx4SVxO6qnyDIEkS5s+fj1OnTiEWi11xe1mWUVtbi0WLFkGno5eB1qARu4HIz8/HHXfcgb179yIUCl30/ixjDMlkEjU1NZgwYQImTZp0HXpKDBea73YDIUkSpk+fDoPBgN27d6OiogKzZs1CYWEhJElCNBpFT08PmpqaUFlZiVtuuYXmK2sUCu4NhiRJmDp1KsrKynDu3Dm89957CIVCGBwcREVFBSZPnozVq1fDYrFQaDXMdQ3uZ+GFwxhDOp2G3W5He3s7zp07h5kzZ2Ly5MnIzc297DHIuC6XC319fejr60N3dzfKyspgNpsv6UqSBJPJhNmzZ2P06NEYGBiA0+nEbbfddkX3r/tss9lw/vx5tLe3c/fZ6XRiYGAgqz7/dbt2ux2dnZ1ob29HeXn5Fd3PGtc1uDfyHNnMc5/ff/99bNmyBfX19XC73VAUBZs3b8acOXPwne98B9XV1Z8KQ8Y9efIkOjo6MHbsWIwbNw5f/vKX0dvbi2PHjmHMmDFYsmTJRV1ZltHY2PgJd8WKFRfcsWPH4s4777xoCNPp9IU+nzhx4kKft2zZgsrKSvz93/89vv71r2fd7sqVK6/Ybsatra3F1q1bP9Hu7373u8u2+5mFMTasn2Qyady2bVsLEyCdTrMXX3xRRGWqqrI//vGPwu4LL7wg5Nrtdnbw4MErbheLxdi//uu/sqKiIgbgoj8mk4lt2LCBDQ0Nfcp97bXX2JEjR1gikWCqqn6i7+l0mp05c4b96U9/Ym63+xO/z8Zta2u7pPvzn/+cFRYWXrbP69evZ06n81Pujh07smrX4/Fc+L2qqiwajbKf/exnV2x348aNzOVyfWLfF2NgYIDV1NRccYwuhtPpZPv37xdy/X4/27Vrl5AbiUTkHTt2PMSyzB19xx0BMp8UTz311GWnsSWTSWzfvh0GgwFbtmxBbm4uFEVBTU0NZs+ejdmzZ3/q00WSJBgMBsyaNQujRo3CwYMHsXbtWpjNZiiKgkOHDl3RnT17NkpLSz/lbt26FU899dRlH86dTCbxyiuvICcnB1u2bIHZbIaqqnjrrbdQWVmJWbNmXbbdUaNG4cCBAxfaZYxh69atePrpp6/Y7ssvvwyTyYTf/OY3n/l7z3Q76CrDGEN7ezs2b96c1dxTVVWxY8cOHDhwAIwx9PT0XHiRX+6UUJIkjBs3DnPnzsXx48fBGEN3dzdycnK43Pr6+k/0OZsn6quqildffRVvvfUWGGPo7OyE2WzGzTfffMV2x48fjzlz5uDEiRNgjMFqtXK1u23bNtTU1NzQX7OygTu44XDY6Pf7C+LxOD1p7BLs3r0bDocj6+3j8ThefPFFKIqC5uZmLFy4MKvvcZIkYfbs2ejv74csy8JuOp3Gzp074XQ6ufr80ksvIZVKobm5GbfffntWEzkkSUJlZSX6+vqQSqXwxhtvcE31SyQS2LZtm9C87BsJ7uC2tLSMe+ihh7758ssvV45Eh7SOqqo4evQo9yfC2bNn4XK5EIvFUFpamrVnNBpRWFh44UmOo0aN4nItFgu8Xi9OnjzJ3efTp0/D5/MhnU6jqKiIq92CggL4/X4cOXKEu91Tp04JL1y4UeD+jrt48WLb008/HViyZElXMpnUHTx48NYTJ05YRApoqaqKhoYG5ObyV+xkjKGhoQH5+fncLvDRonOR0huBQAC9vb2XLGeZTqfR29vLvV+n04lXX30VJpOJawqiJEkwGo14/fXXYTQaua64SpIEvV6P119/XXiZ4s6dOy/7tI3Ltb1r1y50dHRwt+t0OrFr165Lvkl5vV4MDg4KhTsUCqGzs1NojXIsFkNbWxsUReF2k8kkV+Ev7uB2dHRYzGazPHPmzBAALF261Or1emPLly/n3RUURUEwGISIyxiDz+cTcgHA4/EIuU6nE62trfjKV75y0d8rioLnnnuOOwilpaVYuXIl6urqoKpq1uFlH9/7XLFiBY4f51v7wT6+7XTPPffgwIED6Ozs5PILCgqwfPnyC99Xedu+++67sWPHDu4qg6WlpVi2bNklKwwODg6io6MDd911F9d+gY9eF4WFhUKvjWAwCIPBIOTGYjHU1tZmvT13cCdOnBj7zW9+81bm3xaLJWE2m9WCggLeXUGWZZjNZoi4jLHr4kYiEeTm5l7SZYyhqqoK7777Ltd+Z86ciWnTpqGlpQWBQCDrU15ZlhEKhTB16lQ0NTUhGAyiuLg4azcSiaC8vBwLFizAoUOHuPo8d+5cTJ48GadPn0YkEsn6dFlRFEQiEUycOBF33nknPvzwQ67g33LLLZg4ceIlS3BaLJbLjtHliMfjwq6iKMJu5oF+WW/P20B+fr46duxY/gKgF+FGvZG+Zs0ajBkzJuvtc3JysGHDBhgMBlRWVqKxsTHrF3JnZyfKyspgNBpRWVmJhoaGrFzGGM6fP4+ysjLk5OSgurqa67t1Tk4O1q9fD7PZzN3uuXPnLsykWrNmzSVLfF6q3XXr1n3mqzBc19tBN+Il/cxV00ceeSSrx8hIkoSvfe1ruO+++yBJEioqKhAIBNDd3X3Z45P5qnDixAnccccdkCQJM2bMyNr1+/1oaGhAVVUVJEnC3Llz8fDDD2cVCEmSsGrVKnzta1+DJEmYOXMmfD4fenp6smq3sbERixYtgiRJmDdvHle7999/P1auXHnDvulny3UN7o168A0GA3784x/joYceuuyFN71ej69+9at45plnLpxeGQwG3HPPPTh+/Diampogy/InwsA+Lujd19eHvXv3YtmyZReeXvHXrqIoF3V7e3sv6j722GP4h3/4h8tObsi08cwzz1y4MJj5f8eOHUNzc/Ml2+3p6cHevXtx9913X2hXr9fjscceww9+8IMrtnvvvffiv/7rv4QvSN5I0FzlEcJisWDz5s1YsmQJtm7diubmZgSDQTDGYLFYMG3aNDz44IPYtGnTp74b5ufno7q6GkePHsWrr76K8vJylJWVQa/Xw+12o7e3FyaTCatXr76wZO9i7iuvvMLlWiwW/Md//AcWL16M3/72t2htbf1En2+66SZ8+9vfxoMPPoiioqJPuZdrt6enB7m5uZ9qV5IkFBQUYPPmzVi8eDG2bt3K1e5nFZryOEJIkgSz2Yy1a9di5cqV6O7uRltbG86cOYNVq1ahoqICRUVFF70gkVndc9dddyEWi8Fms6GnpweNjY249957sXz5chQWFmbt9vf3o66uDvfdd98V3Uyfv/rVr6K7uxvd3d04fPgwvvWtb12xz2azGV/60pcutGu321FbW4s1a9ZgxYoVV2z3m9/8Ju699150d3ejt7cX7777LtavX3/Zdj+rUHBHmMwjUOfOnYsJEyYgPz8fCxcu5HJnzpyJCRMmIBQKYe7cudxueXk5XC4Xl2uxWDBv3jzMmjULkUhEqM8VFRXo6+sTanfu3Lnw+XxZt/tZg97CCEKDUHAJQoNQcAlCg1BwCUKDcAe3v7+/8J133vl8R0fHxR99zwFd1icIMbivKv/85z+/7YMPPpjx0EMP6f75n//5tKqqUuYGOy+qqkLUzTzCQ7RE5PVodzju9TpWw2lXi30e7viqqjqsPmcLd3AVRWGlpaVeg8FgSiaTut27dy88evRogWhn6+rquD3go4N07Ngx4Xt7x44dg16v5/YCgQD6+/vhdru53UgkgvPnzwsXhWptbRVabpZKpXDq1CnhomCZp2Twoqoqjh07JnRmxRhDXV2d0Bj5fD44HA4MDQ1xu6FQCN3d3fD5fNxuLBbD2bNnEY1Gud1kMsn1WuYO7pIlS3yRSGR8SUmJy2QyqWvWrKlXVTW8fv163l1BlmXo9Xps3LiR2828kB544AEhV1VVIdfhcKC5uRkrVqzgdr1eL44ePYpVq1Zxu+FwGKNGjUJ1dTW3G4/HkZ+fj3Xr1nG76XQaRqMRGzZs4HYzyxNFx4gxJuRmSl0uW7aM23W5XGhoaMDKlSu53UAggNraWqxevZrbjUajePPNN7Penju43/ve95q+973vNfF6BEFcPeiqMkFoEAouQWgQCi5BaBAKLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQCi5BaBB6yiNBaBDu4O7bt++mRx99dFVLS0v2T7G+BDfyUx4JYiThnqu8ffv2CpvNxl577bX58+bNe7e/v3+02+02dnd3czeuKApcLhdEXMYY3G63kAtA2HW73RgaGhJyA4EAnE6nkBuNRoWPVTKZFHZlWdbcGA0NDQmPkc/nE3bD4bDw+MbjcaTT6ay35w7uo48+enLr1q1fMhqN7nQ6Ldnt9hKv12sYGBjg3RVUVYXP54OIm3mSv4gLQNj1+Xxwu91CbigUEnZjsRg8Ho+Qm0ql4PV6hVxFUYY1RqLtAuJj5Ha7hY9zIBAQPs6RSETYTSQSXMsuuYM7evRo/V133WVftmxZk9FoZIsWLTo/MDAQ/+IXv8i7K8iyDJvNBhGXMYa+vj5ht6enR8h1OBzIzc0Vcr1eLxhjQm44HEY0GhVy4/E4PB6PkJtOp2G324VcVVXR399/zcfIZrOhqKhIyHW5XDAYDEJuIBBAKpUSckd8WV9FRYWzoqIi+9LlBEFcdeh2EEFoEAouQWgQuo9LEBqE6uMShAahT1yC0CD0iUsQGoQuThGEBqHgEoQGoe+4BKFBsgpuMpnU19bWft7pdOa/8847N9fX14+7Go3Td1yCECOrKY+KouBXv/rV9Icffjh/x44dY5PJZM7rr7/+uk6nY6qq6hhjUBSFu/FMgSQRF/go+NfaHU6fh+sOp8/X4+/NlHq5HmOk1fHNlqyCm5eXp9x0000OAOaFCxd6+vr6JgLAx0W/bhtu0S/RolDDKfolWlAqGAyir68PHo+H280U/QqFQtzucIp+pdNpNDY2DqvolwjXa4yGW/Srp6cHfr+f243FYrBarX87Rb8SiYQxPz//pra2tmBPT495/Pjxdp1Ox0wmE6OiX9lzPYt+5eXlXZeiX5IkXfMxoqJfH2M2m9P/+Z//+RJ3bwiCGBHoqjJBaBCaOUUQGoQmYBCEBqHgEoQGoeAShAah4BKEBqHgEoQGoeAShAah4BKEBqHgEoQGySq44XDY+Pjjj69tbm6e8MMf/nDNkSNHxox0xwiCuDRZzVUuKChIDwwMuBhj5rKyMs+5c+csd955pxsA+vr6xrhcLmNXVxd345miXyIuY0zYBT6qLyPiut1uOBwOITcQCGBoaEjIjUajcDqdQm6m6JeImyn6JeKqqnpdxmhoaEh4jDIri0TccDgsPL4jVvTLZDIZE4mEubi42JhKpcwAkE6npaGhoSK/328YHBzk7qyqqvD7/RBxGWPCLgBh1+fzwev1CrmhUEjYjcVi8Pl8Qm4qlRJ2M0W/tDRGbrcbHo9HyA0EAsJjFIlEhI/ViBT9SiaThmXLlqmDg4OW4uJiddSoUfkAYDQa2e23397Z398fX7p0KXdnZVnGwMAARFzGGHp7e4Xd7u5uIdfhcMBsNgu5Xq8XqqoKueFwGJFIRMiNx+Nwu91CbjqdxuDgoJCrqir6+vqu+RjZbDYUFhYKuS6XC3q9XsgNBAJIJpNC7ogs6zOZTPK3vvWt97h7QxDEiEBXlQlCg1BwCUKD0EJ6gtAgtJCeIDQInSoThAah4BKEBqHgEoQGoeAShAah4BKEBsk6uNFo1CzLsi4SiZhjsRh/XQiCIK4aWQU3GAzmVFdXbzh27FjF448//n8effTRu0e6YwRBXJqs5ioXFRWlysvLraNHj05VV1dbn3322dkAkEqldDU1NfNOnTqVb7FYuBtXVRWnTp1CYWEht8sYw6lTp1BcXMztAsDp06dRUlLC7fn9fvT09CCZTHK74XAYVqtV6P51PB7H6dOnYTQaud1UKoXGxkbk5eVxu4qioLGxEQUFBdxuZoyKioq4XUB8jLxeL2w2G2KxGLcbDAZx/vx5oQJp0WgUra2tQkXOkskkV7GwrIKbSqX0kUik/P333y94/vnn5z3yyCNvAkBOTo66cOHCLrvdHl+yZAl3ZxVFgdPphIjLGIPdbhdygY9WkIi4LpcL+fn5Qq7f74dOpxNyI5EIEomEkJtMJhEMBoVcWZbhdruFXFVV4XA4hMdocHBQyLXb7SgpKRFyPR4PcnJyhNxQKARFUYTcWCyGDz/8MOvtswquXq9Xf/nLX36gKApbvnz5GbPZfOGtobi4OJqfn6+OGjWKu7OyLCM/Px8iLmPsurjJZBIFBQXC7VosFiHXaDQKu/F4XPjvTafTwq6qqsMao7y8PCE3FosJHytZloXHV6fTCbdrMpm4SopmG1w2depUO3dvCIIYEeh2EEFoEAouQWgQCi5BaBAKLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQCi5BaJCsgptKpXT19fU3ezweS319/czm5ubRV6NxesojQYiR1ZTHdDot/fSnP533b//2b8q5c+eK9+/ff+uePXteBQDGmPTxf7kbZ4xd+BFxRdv9632IeNe6XS3+vcMZ37/ch6hzPcZ3uH9vtmQV3Pz8fGXGjBm2kpISuaKiwitJ0lQASCaTujfeeOP2Dz/8sEBkGZSqqqirqxMenKNHj3J7f+mKfOIHAgEMDAzA5XJxu5FIBOfPn4ff7+d2E4kEWltbuZZ+ZUilUjh16hRSqRS3qygK6uvroaoqt5sZXxGGM0aZint2O//0+lAohJ6eHni9Xm43FovBarUiHA5zu8lkkms5YFbBTSQSRoPBULF3796xOp1OmT59uhcATCaTWl1dfZwxFl6/fj13Z2VZhsFgwMaNG7ldxhgkScKDDz4o5AIQch0OB5qbm7FixQpu1+v14ujRo1i1ahW3Gw6HcejQIVRXV3O78XgcFosF69at43bT6TRycnKwYcMGbldVVej1ejzwwAPc7nDGyGazwWq1YtmyZdyuy+VCQ0MDVq5cye0GAgHU1tZi9erV3O6IFP0ym83pX//61y9mDqZer6cnmRPEdSTr+rg6nY7CShB/I1DtIILQIFQ7iCA0CE3AIAgNQsElCA1C33EJQoPQd1yC0CB0qkwQGoSCSxAahIJLEBokq+BGIhHjL37xi/udTueompqaab/97W8XjHTHCIK4NFlNeTSZTPKZM2eCbW1tY3/961/PUxQl+fDDDzcCgMPhKPb7/ZLNZuNeeqIoCnw+n2Sz2bivUjHG8HG7Qle4RF2XywWPxyPkBgIBeL1eITcajQofq2QyKezKsjysMRJ1AfExcjgcwmPk9XqF3XA4LDy+8Xhc5Vlhl1VwjUYjs1gseO+99yaWlpZO6Ojo0Pt8Pl1BQQHr7OwcNzg4+LLVaj3F21lVVXV2u32+1Wo9yesyxiS73f4Fq9XawOsCgN1uv03E9fl8hQMDAxOsVus5XjccDucPDAx8zmq1tvG68XjcbLPZZlqt1mZeN5VKGQcHByutVutpXldRFL3dbr/FarU28rrDHCNh1+PxFPf394+1Wq0dvG4gELAMDAyUW63Ws7xuNBrNtdls061Wayuvm0gkclKpVE7Wwl8udr7UTzweNzz//PMr9uzZc3soFMr985//PC3zu1QqpX/zzTfnZrOfv/6RZVnat2/f50Vcxhj27Nlzy7V23W635ciRI9NF3EAgkFtbWztTxI1EIjk1NTWzRdxEImE4cOCA0Bil02nd/v3754ke5717917z8R0aGiqsq6ubJuJ6vd68w4cPzxBxQ6GQ6Z133pkl4sZiMeNbb701J9vts/qOazab5U2bNr21atWq+oKCgvh9993XnfldbW3tZLfbnR+NRrmr1NfX13+usLBQaBbGzp07bwoEAnl+vz/7d6mP2b1792S/32/xeDxmXjc3Nzfa1dVVGAwGuYvNulyuMQ6HY2IoFMrndd1ut85ms5UcOXJkDK975syZYo/HM/69996bweu+//77n3O5XIW9vb3cx/mDDz6Y6nK5coPBYNar0BKJhGHbtm1Le3p6Svx+f87evXunZeumUind9u3b74xGo6YzZ86UNTU1lWbryrIsvfrqq3f4fL6i/v7+UTt37pyZrauqqrRz587bHQ7HmHA4nLtjx47KVCqV9ev6z3/+8wKbzTZ+9uzZvdu2bZufjTOsq8o+n0+/efPm2w4dOlR47NixCby+0+kMv/HGG1kfoL9k1KhR7nfffbfk4MGDY3ndWCwW3LJly/Ta2trxvO6777476Sc/+cmXXS4Xd7XmkydP3nT48OH8VCrF/TiJZ599dv6+ffvG9/b2crszZswItLS0yN3d3VzVx6PRqPTTn/500b//+7/P6+3t5aowHYvFdD/72c9u/8Mf/jD20KFDWYdPkiT1jTfeyGtubq6sq6srfemll25TFCWrEEiSxPbu3ZvT398/rru7O+ftt98el227BoOB1dTU6M6fPz9h1KhR3S+88MKCWCyW1RuOTqdjH3zwgdra2jrJYDCEfve731UlEoms36zq6+vTdXV10/77v/97/i9+8Ys7PB7PFf/eYQWXMSbl5OSwMWPGxBRF4S6VPmXKlLAkSYpI2/39/fnRaNT0la98hfsZMpWVlYl58+YNhsNhE6/7wQcffGHixImF/f39Wb8oMowbN+6U0WgM/e///u9cXjcSiRRMmDAh2dDQ8AVe12AwKGfOnBnzxS9+sZ3HUxRFMhqNqdGjRztSqRTXGUZeXp76/e9//7TJZCpLp9NZv4hNJpM6YcKEIGPMMGnSpHhZWVnWbRqNRjZp0qSg2WxWb7/99hBPfwFg8uTJAb1ez1pbW6eXl5d35uXlZX21aMqUKQGdToeKigrvx/Wjs87WlClTgs3NzXktLS2ft1gsExwOxxXPBIcV3NLSUnn58uVdOTk5JfPnz7fx+qFQaGpxcXFJb28v96dXT09P5eLFi6VQKMR9uvvee+9VjB8/PnfJkiVOXvepp57a+6Mf/WjH7Nmz+3jdjo6OUUVFRSX33HNPL6/7gx/8oLGkpMSwaNGi87yuz+fT33nnnb1TpkyJ8XiFhYXq3/3d3w0sWLBAHTt2rJu33XHjxumWLFnSt3Tp0q5snWQyaZg0adKk3t5eUyQSsUydOrUr2yeupNNp/fjx4yc3NjZO7erqmmyxWKaqqprVp7Usy7rS0tIpbW1t02w225zFixcbZVnOylVVVbJYLJ/r6uqafODAgYVVVVV2o9GYzsYFAJPJ9LlJkyYVbN++/YXvfOc7e6dMmZK4kiMxRvOFCUJr0MwpgtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIP8PnxFgaq7ZgpQAAAAASUVORK5CYII="
      },
      "metadata": {
       "needs_background": "light"
@@ -776,14 +751,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [8, 7] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        8,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -804,10 +779,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -818,7 +791,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"Alice is trying to build a vertical line again. I need to block her next move otherwise she will have a chance to win. I will place my piece at [9, 7].\",\n",
       "    \"move\": [\n",
       "        9,\n",
@@ -851,13 +824,17 @@
     "            msg = player(msg)\n",
     "            \n",
     "            i += 1"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:10:33.989202Z",
+     "start_time": "2024-03-27T08:09:59.446824Z"
+    }
+   }
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "## Future Direction\n",
     "\n",
@@ -868,7 +845,10 @@
     "- Fine tune a model to improve the performance of the agent\n",
     "\n",
     "For complete code, we provide in [code/game_gomoku.py](./code/game_gomoku.py)."
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   }
  ],
  "metadata": {

From a0882fa26db5518552ade0fe13a6d2ea5b3ee749 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 22 May 2024 20:44:36 +0800
Subject: [PATCH 28/55] revert back unnecessary changes

---
 examples/game_gomoku/main.ipynb | 260 +++++++++++++++++---------------
 1 file changed, 140 insertions(+), 120 deletions(-)

diff --git a/examples/game_gomoku/main.ipynb b/examples/game_gomoku/main.ipynb
index 7044a8243..569321c15 100644
--- a/examples/game_gomoku/main.ipynb
+++ b/examples/game_gomoku/main.ipynb
@@ -31,6 +31,13 @@
   {
    "cell_type": "code",
    "execution_count": 1,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:57.602994Z",
+     "start_time": "2024-03-27T08:09:57.565787Z"
+    },
+    "collapsed": false
+   },
    "outputs": [],
    "source": [
     "YOUR_MODEL_CONFIGURATION_NAME = \"{YOUR_MODEL_CONFIGURATION_NAME}\"\n",
@@ -39,38 +46,38 @@
     "    \n",
     "    # ...\n",
     "}"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:57.602994Z",
-     "start_time": "2024-03-27T08:09:57.565787Z"
-    }
-   }
+   ]
   },
   {
    "cell_type": "markdown",
+   "metadata": {
+    "collapsed": false
+   },
    "source": [
     "## Step 1: Prepare a board for Gomoku\n",
     "\n",
     "First we create a `BoardAgent` class by inheriting from `AgentBase`, which manage the game board and update the game status as follows.\n",
     "\n",
     "To create a better visual experience, we also provide a function `board2img` to convert the board to an image. "
-   ],
-   "metadata": {
-    "collapsed": false
-   }
+   ]
   },
   {
    "cell_type": "code",
    "execution_count": 2,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:58.153932Z",
+     "start_time": "2024-03-27T08:09:57.571266Z"
+    },
+    "collapsed": false
+   },
    "outputs": [],
    "source": [
     "import numpy as np\n",
     "import matplotlib.pyplot as plt\n",
     "import matplotlib.patches as patches\n",
     "\n",
-    "def board2img(board: np.ndarray, output_dir: str)->str:\n",
+    "def board2img(board: np.ndarray, save_path: str)->str:\n",
     "    size = board.shape[0]\n",
     "    fig, ax = plt.subplots(figsize=(10, 10))\n",
     "    ax.set_xlim(0, size - 1)\n",
@@ -100,29 +107,30 @@
     "    ax.set_xticklabels(range(size))\n",
     "    ax.set_yticklabels(range(size))\n",
     "    ax.invert_yaxis()\n",
-    "    plt.savefig(output_dir, bbox_inches='tight', pad_inches=0.1)\n",
+    "    plt.savefig(save_path, bbox_inches='tight', pad_inches=0.1)\n",
     "    plt.close(fig)  # Close the figure to free memory\n",
-    "    return output_dir"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:58.153932Z",
-     "start_time": "2024-03-27T08:09:57.571266Z"
-    }
-   }
+    "    return save_path"
+   ]
   },
   {
    "cell_type": "markdown",
-   "source": [
-    "The following code shows the implementation of the `BoardAgent` class. The agent manages the game board and updates the game status."
-   ],
    "metadata": {
     "collapsed": false
-   }
+   },
+   "source": [
+    "The following code shows the implementation of the `BoardAgent` class. The agent manages the game board and updates the game status."
+   ]
   },
   {
    "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.392069Z",
+     "start_time": "2024-03-27T08:09:58.159470Z"
+    },
+    "collapsed": false
+   },
    "outputs": [],
    "source": [
     "import numpy as np\n",
@@ -229,18 +237,13 @@
     "            else:\n",
     "                count = 0\n",
     "        return False\n"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.392069Z",
-     "start_time": "2024-03-27T08:09:58.159470Z"
-    }
-   },
-   "execution_count": 3
+   ]
   },
   {
    "cell_type": "markdown",
+   "metadata": {
+    "collapsed": false
+   },
    "source": [
     "### Step2: Prepare a Gomoku Player Agent\n",
     "\n",
@@ -250,13 +253,18 @@
     "2. Within the agent, to enable the agent to think, we ask LLMs to respond in a dictionary format, which contains the thought and the move (\"thought\" field must come before \"move\"). To achieve this, we prepare a parsing function to extract the dictionary from response. \n",
     "\n",
     "The implementation of the Gomoku agent is as follows: "
-   ],
-   "metadata": {
-    "collapsed": false
-   }
+   ]
   },
   {
    "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.397713Z",
+     "start_time": "2024-03-27T08:09:59.392729Z"
+    },
+    "collapsed": false
+   },
    "outputs": [],
    "source": [
     "import json\n",
@@ -327,15 +335,7 @@
     "        \n",
     "        # Hide thought from the response\n",
     "        return Msg(self.name, response[\"move\"], role=\"assistant\") \n"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.397713Z",
-     "start_time": "2024-03-27T08:09:59.392729Z"
-    }
-   },
-   "execution_count": 4
+   ]
   },
   {
    "cell_type": "markdown",
@@ -346,17 +346,25 @@
   },
   {
    "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.446367Z",
+     "start_time": "2024-03-27T08:09:59.398116Z"
+    },
+    "collapsed": false
+   },
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m2024-03-27 16:09:59.427\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models\u001B[0m:\u001B[36mread_model_configs\u001B[0m:\u001B[36m171\u001B[0m - \u001B[1mLoad configs for model wrapper: post_api\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.435\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m_create_monitor_table\u001B[0m:\u001B[36m341\u001B[0m - \u001B[1mInit [monitor_metrics] as the monitor table\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.435\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m_create_monitor_table\u001B[0m:\u001B[36m342\u001B[0m - \u001B[1mInit [monitor_metrics_quota_exceeded] as the monitor trigger\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.436\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m311\u001B[0m - \u001B[1mSqliteMonitor initialization completed at [./runs/run_20240327-160958_kbf2ta/agentscope.db]\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.440\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models.model\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m257\u001B[0m - \u001B[1mInitialize model [post_api]\u001B[0m\n",
-      "\u001B[32m2024-03-27 16:09:59.441\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models.model\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m257\u001B[0m - \u001B[1mInitialize model [post_api]\u001B[0m\n"
+      "\u001b[32m2024-03-27 16:09:59.427\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models\u001b[0m:\u001b[36mread_model_configs\u001b[0m:\u001b[36m171\u001b[0m - \u001b[1mLoad configs for model wrapper: post_api\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.435\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m_create_monitor_table\u001b[0m:\u001b[36m341\u001b[0m - \u001b[1mInit [monitor_metrics] as the monitor table\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.435\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m_create_monitor_table\u001b[0m:\u001b[36m342\u001b[0m - \u001b[1mInit [monitor_metrics_quota_exceeded] as the monitor trigger\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.436\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m311\u001b[0m - \u001b[1mSqliteMonitor initialization completed at [./runs/run_20240327-160958_kbf2ta/agentscope.db]\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.440\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models.model\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m257\u001b[0m - \u001b[1mInitialize model [post_api]\u001b[0m\n",
+      "\u001b[32m2024-03-27 16:09:59.441\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models.model\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m257\u001b[0m - \u001b[1mInitialize model [post_api]\u001b[0m\n"
      ]
     }
    ],
@@ -381,18 +389,13 @@
     ")\n",
     "\n",
     "board = BoardAgent(name=\"Host\")"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.446367Z",
-     "start_time": "2024-03-27T08:09:59.398116Z"
-    }
-   },
-   "execution_count": 5
+   ]
   },
   {
    "cell_type": "markdown",
+   "metadata": {
+    "collapsed": false
+   },
    "source": [
     "### Step 3: Start the Gomoku game\n",
     "\n",
@@ -401,26 +404,32 @@
     "In this game, we use a message hub to share messages between the two players (board agent doesn't have memory), so that one player can hear what the board agent says to the other player, and what moves the other player makes. \n",
     "\n",
     "Note here we only show 10 steps of the game. You can adjust the `MAX_STEPS` to play more rounds."
-   ],
-   "metadata": {
-    "collapsed": false
-   }
+   ]
   },
   {
    "cell_type": "code",
    "execution_count": 6,
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:10:33.989202Z",
+     "start_time": "2024-03-27T08:09:59.446824Z"
+    },
+    "collapsed": false
+   },
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: Welcome to the Gomoku game! Black player goes first. Please make your move.\n"
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: Welcome to the Gomoku game! Black player goes first. Please make your move.\n"
      ]
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAe5UlEQVR4nO2da5BU1bn3n75Nd0/3DDPcQQQEBggiHiEKqOAxr0RS1ouHypATIppLqbGiRYwfktIylUvpm1SSDzExSSXlJQmGBC9RbqKojCgwDMMAM8PQMMPcunu6Z/q6u/fuvXvv3nut9wN0XuvUyWuvpxnm7Mrzq5qyLPq31+r99L8vu9fqx8E5B4Ig7IVzoidAEIQ4FFyCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAhwsE9d+5cw44dO9aOjo7WjseECIL4dByiCzC2bt26obGx0Zw9e7b69NNPt+m67uacu8ZpfgTxL4Pb7TbdbrdV0W1FD26apnvp0qWJRCIxSdd15969ez97+vTpZ5YvXz5L9FiMMTh79iysWLFCVAXOOXR1dcGNN94o7AIAdHZ2olxFUWBsbAwWLlwo7GqaBpFIBBYvXizs6roOAwMD8JnPfEbYNU0TLly4ANdff72wyxiDnp4euOGGG4Rdzjl0d3df9frm83lIpVKwYMECYbdQKEAsFoOmpiZht1gswtDQECxdulTYLZVK3O/3/58tW7bsrEjgnAv9vfrqq/MffPDBL7W3t0/jnIOu655XXnmliyMolUr8z3/+M0bljDH+xz/+Ee2+/PLLKDcWi/EDBw6g3FQqxXfv3o1y8/k8f+2111Cuqqp8586dKNcwDL5jxw6Ua1kW/9Of/oRyq6lRJBLhBw8eRLljY2N83759KDebzfK///3vKFdRFHPXrl2P8ApzKPyKu2XLlqEtW7YMiXoEQVw56KoyQdiQCQ2uw+GYyOEJwrZMaHA5bSkkCBT0VpkgbAgFlyBsCAWXIGwIBZcgbAgFlyBsCAWXIGwIfY9LEDZEOLhHjhy59umnn77vSgxO3+MSBA7htcrz5s2Lh0IhxhhzMMbg3Llzc8LhsPfMmTPCgzPGIBwOA8blnKNdAIBIJIJyU6kUDAwMoNxcLgeDg4MoV1VVGBoaQrmGYcDw8DDKtSwL7U5UjcbGxtDnKpvNouurKAq6vsViEXRdr/j2wsF1OBweXdfrc7mcr66ursgYc3DOgTEmeihgjP3jDwN23E+OjRkTO241LmPMdm4197fs/yudZxGEg+t0On0PP/xwbzqdrm1sbNSWL18eCYVC+sqVK0UPBaZpQigUAozLL+/1xLpdXV0oNx6Pg2VZKDedTkM+n0e5sizD6OgoytU0DQYHB1FuqVSCCxcuoNzyXt6rXaNoNAoulwvlJhIJKBaLKFeSJEin0yi3UChANBqt+PbCwZ09e3Z206ZNLaIeQRBXDvo6iCBsCAWXIGwIBZcgbAgFlyBsCK2cIggbQr+AQRA2hN4qE4QNoeAShA2h4BKEDREObjab9SYSiUmmadKVJYKYIISXPL7wwguzQqHQzevXrz/zta99rW88JkUQxP8f4eA+8MADkccee+z6+vp6tdz06/jx40GXS7xhH2MMjh8/Dh6PR9jlnENbWxt4vV5hFwCgra0NfD6fsCdJEgwPD4MkScKuoihw4cIFUFVV2NU0Dbq6usA0TWHXMAw4efIk6us3y7KgtbUV3G7hh8o/6ltTUyPsluuLqVE6nYZYLAbpdFrYzefz0N/fD7IsC7uqqkJPT4/Q9rwyuq4L1Va4Gs8+++wtfr9//rJly1q9Xi/7/Oc/36koSuHee+8VPRRYlgWapgHG5ZyDoigoF+DSbhuMOzo6Ct3d3bBhwwZhN5PJQENDA9xzzz3CrizL4PV6UXMuFosAACjXNE3QdR3lMsZAVdWrXt+RkRG4cOECfO5znxN2k8kkdHR0wMaNG4XdXC4HgUAANm3aJOyqqgrvvfdexbcXDu6PfvSjM4ZhXAgEAjIAgM/nK3k8Hu73+0UPBaZpQk1NDWBcznlVrsfjQbk+nw89bjVuNecKAND3t1QqocdljP1L1UjX9arOldNZ+SUn4eA2NjZqAKCJegRBXDno6yCCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAhwsE9evTo1H379q1KpVLiq78JgrgiCC95HBgY0F988cXFX/rSl4rf+ta3ekzTdDHGULtWLMsCy7JQLucc7QJcWht6tedcjWuaZlXjYu9vNeMyxqqqbzVztmN9RX6DTTi4a9as0drb2xOqqnoub+tb1draWid6HID/t+0LsyUQ4NLWPMyWQABAbyfM5XIQDochk8kIu4VCAXp7e0FRFGG3WCxCd3c3GIYh7JZKJejo6EA3ozp+/LjQAvgynHM4fvw4aksgAL6+2WwWYrEYJBIJYVeWZRgYGIBcLifsapoG586dA00TX8ovWlfhM7p3797rXC5X49q1a0Ner5dt3rz5hGma8n33ibfMNU0TnE4n3H///cJu+RkZ61qWhXLj8Th0dnaitn2l02k4evQoatuXLMvQ2NgIzc3Nwq6maeD3+2Hr1q3CbqlUApfLBdu2bRN2y08UV7tG0WgUQqEQautlIpGA9vZ21NZLSZKgpaUFNm/eLOwWCgXYv39/xbcXDu4TTzzRBwBX5Jcv6HeVCQIH/a4yQdgQ+jqIIGwIBZcgbAj1DiIIG0KfcQnChtBbZYKwIRRcgrAhFFyCsCEUXIKwIXRVmSBsiHBwc7mc8/HHH//f0Wh0SrWD01VlgsAhvFZ59+7di/7617/OffDBBztnzZqV6e/vnxGPxz3nz58XHpwxBvF4HDAu5xztAlzqAYRxU6kURKNRlJvL5WBkZATlFgoFiMViKFfXdfS5sizLdjUaGxtD1yiTyaBdWZbR9dU0TWiHkHBwJUmas2jRommDg4OzlixZEpEkKVAoFFyYbW6MMVAUBbVFDuDSgxnrYseVJAlkWUa5+Xwe7Wqahp6zYRho17IstMs5r6pGWLeaGlXjFgoF9LkqFotC2y6Fg7t9+/ZDN998c8/SpUsVj8fDV61aNTAwMFC89dZbRQ8FpmnC4OAgYFzOOfT19aHd3t5elBuPx8HtdqPcdDoNhmGgXFmWIZvNolxN0yAej6PcUqkEw8PDKJcxBv39/Ve9RtFoFGpra1FuIpEAzjnKlSQJFEVBuYVCAVKpVMW3R+1wXrt27RjGIwjiykBfBxGEDaGvgwjChtAmA4KwIfRWmSBsCAWXIGwIBZcgbAgFlyBsCF1VJggbQleVCcKGCAd3eHh41oEDB9YoiuIfjwkRBPHpCC95/Oijj65vb2/3rV27lgMAXG765SiVSsKDl5skYdxyiwqMWx4b45abYJH76ZSbfmFrxBiz1f2t1h3Xpl9z5sw5feTIkZUvvvjisscee+zM5aZfQczn1XJDKUxTKM45tLW1QU1NjbALAHDixAnwer3CniRJEA6HIZvNCruKokBvby8UCgVht1gsQldXF+pBUSqV4OTJk8IewKUnOGxjtnKNsI3ZsPXNZDIQi8WEFu2XyefzMDAwAPl8XthVVRXOnTsHxWJR2NV1Xej2wok5c+ZMQyAQmLR27dqeK9H0y+FwoBpKlZt+YV3TNFHuRDb9amhoQDf98vl8E9L0i3N+1WtETb/+G77zne/0A0C/qEcQxJWDvsclCBtCwSUIG0LBJQgbQsElCBtCwSUIG0LBJQgbQsElCBtCwSUIG0LBJQgbQsElCBsiHNwPP/xw2ve+971Nb7755vTxmBBBEJ+O8FrlX/3qV2scDof/pptucliW5bh48eLMWCxWEwqFhAe3LAtisRhgXM452gW4tFkA4yaTSQiHwyhXkiSIRCIot1Ao/GPxvCi6rsPIyAjKNU0TfZ4ZY1XVF1uj0dFRdI0ymQy6RrIso2s07k2/5s6dqwNA8ezZs9c3NzcnZFn2a5rmzOVyoocCxhioqgoYl3MOmqahXABAu7Iso+dcjauqKnrOhmGgXcuyqEYVoigK2r0aTb9aT548eX1TU1PI7Xbzm266afDixYvFNWvWiB4KTNOE/v5+wLjlplBY9/z58yg3Ho+Dy+VCuel0GorFIsotd5DDuJqmwcjICMotlUowODiIchlj0NfXd9VrFI1Gwe/3o9xEIgGWZaHccqc/jFsoFCCRSFR8e+HgLliwQF6wYMFxUY8giCsHXVUmCBtCwSUIG0LBJQgbQsElCBtCnQwIwoZQJwOCsCH0VpkgbAgFlyBsCAWXIGyI8MopWZY9pmn6fD5f0e/345rCEARRFcKvuF1dXTMeeeSR//zLX/6yfDwmRBDEpyP8invbbbdFf/azn0nr1q3r13XdeeDAgZtOnDgRxDTQYoxBe3s7+P3iHTs559De3g6BQEDYBQA4efIkBINBYU+SJBgaGgJFUYRdRVHg/PnzQtu3ymiaBp2dncIewKWNAidOnEA137IsC06cOAE+n0/YLdeotrZW2AXA1yidTsPIyAhql04+n4eLFy+CpmnCrqqqcPbsWbAsS9jVdV2o8ZdwcHt7e4M+n89csmRJHgBg/fr1oXQ6rd59992ihwLLsiCXywHG5ZxDJpNBuQAAqVQK5Y6NjUF3dzfcddddwm4mkwG/348at/xEgXGLxSIUi0WUa5omyLKMchljkM1m0fXF1mhkZAR6e3vhzjvvFHZTqRTU19ejxs3lcuB2u1GuqqrQ0tJS8e2Fg3vNNdeov/71r98p/38wGCz6fD5WV1cneigwTRN8Ph9gXM75hLiKooDf70e5hmGgXQBAu263G31/S6US2mWMTUiNgsEg+lxpmoZ2LctCu06nE5zOyj+5Cgc3EAiwQCAg3gD0v4FWThEEDlo5RRA2hNYqE4QNoVdcgrAhtHKKIGwIBZcgbAgFlyBsCAWXIGwIBZcgbAgFlyBsiHBww+Fw/fvvv39jb28vbuX4J6DvcQkCh/CSxx/84Ac3Hz58ePEjjzzi/O53v3uaMebgnAv1PSnDGAOsyzlHu2X/ao9bjTtR56qace0452rryxiras6VIhxcy7L4lClT0m6326vruvPNN9+85ejRo3XYyba2tgp7AJdO0rFjx4QWZn+SY8eOgcvlEvYkSYJwOAzJZFLYVRQF+vr60E2huru7UdvNDMOAU6dOgWmawq5lWdDW1oZaLMMYg2PHjqHeWXHOobW1FVWjTCYD8XgcRkdHhd18Pg8DAwOQyWSEXVVV4dy5c1AoFIRdXdfHd5PBunXrMoqizGxsbEx4vV72xS9+sY0xJt93332ihwLTNMHlcsH9998v7JYfSA888ADKZYyh3Hg8Dp2dnbBx40ZhN51Ow9GjR2HTpk3CrizLMHnyZGhubhZ2NU2DQCAAW7duFXZLpRJ4PB7Ytm2bsMsYA6fTia4R5xzllltdbtiwQdhNJBLQ3t4O99xzj7ArSRK0tLTA5s2bhd1CoQD79++v+PbCwX3ooYfOPPTQQ2dEPYIgrhx0VZkgbAgFlyBsCAWXIGwIBZcgbAgFlyBsCAWXIGwIBZcgbAgFlyBsCAWXIGwI/cojQdgQ4eDu3bt34fbt2zd1dXU1Vjs4/cojQeAQXqu8c+fOpmg0yl999dWVK1as+CAcDk9NJpOegYEB4cEty4JEIgEYl3MOyWQS5QIA2k0mkzA6OopyJUmCsbExlFsoFNDnStd1tGuapu1qNDo6iq5RJpNBu7Iso+uraRqUSpV3rRUO7vbt20/+7ne/+5zH40mWSiVHLBZrTKfT7kgkInooYIxBJpMBjFtu+oVxAQDtZjIZSCaTKDefz6NdVVUhlUqhXMMwIJ1Oo1zLsqqqEXZcAHyNkskk+jxLkoQ+z4qioN1isSi07VI4uFOnTnXdeeedsQ0bNpzxeDx8zZo1fZFIRLvjjjtEDwWmaUI0GgWMyzmH4eFhtDs4OIhy4/E4+P1+lJtOp4FzjnJlWYZCoYByNU2DVCqFckulEsRiMZTLGINwOHzVaxSNRmHSpEkoN5FIgNvtRrmSJIFhGCh33Lf1NTU1jTU1NY2JegRBXDno6yCCsCEUXIKwIfQ9LkHYEOrWRxA2hF5xCcKG0CsuQdgQujhFEDaEgksQNoQ+4xKEDakouLquu1paWm4cGxsLvP/++0vb2tpmXInB6TMuQeCoaMmjZVnwk5/8ZNGjjz4a2LVr13Rd12tee+2115xOJ2eMOTnnYFmW8ODlBkkYF+BS8K+2W82cq3WrmfNE3N9yq5eJqJFd61spFQW3trbWWrhwYRwAfLfccktqeHj4GgCAy02/bq626Re2KVQ1Tb+wDaVyuRwMDw9DKpUSdstNv/L5vLBbTdOvUqkEHR0dVTX9wjBRNaq26dfg4CBks1lhV1VVCIVC/3OafhWLRU8gEFh49uzZ3ODgoG/mzJkxp9PJvV4vp6ZflTORTb9qa2snpOmXw+G46jWipl+X8fl8pV/84hc7hGdDEMS4QFeVCcKG0MopgrAhtACDIGwIBZcgbAgFlyBsCAWXIGwIBZcgbAgFlyBsCAWXIGwIBZcgbEhFwZVl2fPkk09u6ezsnP3tb3/7i0eOHJk23hMjCOKfU9Fa5bq6ulIkEklwzn2zZs1KnT9/Pnj77bcnAQCGh4enJRIJT39/v/Dg5aZfGJdzjnYBLvWXwbjJZBLi8TjKlSQJRkdHUW6hUICxsTGUW276hXHLTb8wLmNsQmo0OjqKrlF5ZxHGlWUZXd9xa/rl9Xo9xWLR19DQ4DEMwwcAUCqVHKOjo5Oy2ax7ZGREeLKMMchms4BxOedoFwDQbiaTgXQ6jXLz+TzaVVUVMpkMyjUMA+2Wm37ZqUbJZBJSqRTKlSQJXSNFUdDnalyafum67t6wYQMbGRkJNjQ0sMmTJwcAADweD1+9evXFcDisrV+/XniypmlCJBIBjMs5h6GhIbQ7MDCAcuPxOPh8PpSbTqeBMYZyZVkGRVFQrqZpkEwmUW6pVIKRkRGUyxiD4eHhq16jaDQK9fX1KDeRSIDL5UK5kiSBrusod1y29Xm9XvPLX/7yIeHZEAQxLtBVZYKwIRRcgrAhtJGeIGwIbaQnCBtCb5UJwoZQcAnChlBwCcKGUHAJwoZQcAnChlQc3EKh4DNN06koik9VVfG+EARBXDEqCm4ul6tpbm7eduzYsaYnn3zy37dv3/758Z4YQRD/nIrWKk+aNMmYO3duaOrUqUZzc3Po+eefXwYAYBiG8+DBgytOnToVCAaDwoMzxuDUqVNQX18v7HLO4dSpU9DQ0CDsAgCcPn0aGhsbhb1sNguDg4Og67qwK8syhEIh1PfXmqbB6dOnwePxCLuGYUBHRwfU1tYKu5ZlQUdHB9TV1Qm75RpNmjRJ2AXA1yidTkM0GgVVVYXdXC4HfX19qAZphUIBuru7UU3OdF0XahZWUXANw3ApijL3ww8/rHvppZdWPPbYY/sBAGpqatgtt9zSH4vFtHXr1glP1rIsGBsbA4zLOYdYLIZyAS7tIMG4iUQCAoEAys1ms+B0OlGuoihQLBZRrq7rkMvlUK5pmpBMJlEuYwzi8Ti6RiMjIyg3FotBY2Mjyk2lUlBTU4Ny8/k8WJaFclVVhY8//rji21cUXJfLxZ599tnDlmXxu+++u8fn8/3jqaGhoaEQCATY5MmThSdrmiYEAgHAuJzzCXF1XYe6ujr0uMFgEOV6PB60q2ka+v6WSiW0yxirqka1tbUoV1VV9LkyTRNdX6fTiR7X6/UKtRStNLh8/vz5MeHZEAQxLtDXQQRhQyi4BGFDKLgEYUMouARhQyi4BGFDKLgEYUMouARhQyi4BGFDKLgEYUMqCq5hGM62tralqVQq2NbWtqSzs3PqlRicfuWRIHBUtOSxVCo5nnrqqRU//OEPrfPnzzfs27fvpt27d/8NAIBz7rj8X+HBOef/+MO42HH/6zEw3tUe1473t5r6fvIYWGci6lvt/a2UioIbCASsxYsXRxsbG82mpqa0w+GYDwCg67rzjTfeWP3xxx/XYbZBMcagtbUVXZyjR48Ke590Ma/4kiRBJBKBRCIh7CqKAn19fZDNZoXdYrEI3d3dQlu/yhiGAadOnQLDMIRdy7Kgra0NGGPCbrm+GKqpUbnjXiwmvrw+n8/D4OAgpNNpYVdVVQiFQiDLsrCr67rQdsCKglssFj1ut7tpz549051Op7Vo0aI0AIDX62XNzc3HOefyfffdJzxZ0zTB7XbD/fffL+xyzsHhcMBXv/pVlAsAKDcej0NnZyds3LhR2E2n03D06FHYtGmTsCvLMrz77rvQ3Nws7GqaBsFgELZu3SrslkolqKmpgW3btgm7jDFwuVzwwAMPCLvV1CgajUIoFIINGzYIu4lEAtrb2+Gee+4RdiVJgpaWFti8ebOwOy5Nv3w+X+m55577c/lkulwu+iVzgphAKu6P63Q6KawE8T8E6h1EEDaEegcRhA2hBRgEYUMouARhQ+gzLkHYEPqMSxA2hN4qE4QNoeAShA2h4BKEDakouIqieJ555pn/GBsbm3zw4MEFv/nNb1aN98QIgvjnVLTk0ev1mj09PbmzZ89Of+6551ZYlqU/+uijHQAA8Xi8IZvNOqLRqPDWE8uyIJPJOKLRqPBVKs45XB4XdYUL6yYSCUilUihXkiRIp9Mot1AooM+Vruto1zTNqmqEdQHwNYrH4+gapdNptCvLMrq+mqYxkR12FQXX4/HwYDAIhw4dumbKlCmze3t7XZlMxllXV8cvXrw4Y2Rk5C+hUOiU6GQZY85YLLYyFAqdFHU5545YLPbZUCjULuoCAMRisZsxbiaTqY9EIrNDodB5UVeW5UAkErkuFAqdFXU1TfNFo9EloVCoU9Q1DMMzMjKyPBQKnRZ1LctyxWKxfwuFQh2ibpU1QrupVKohHA5PD4VCvaKuJEnBSCQyNxQKnRN1C4WCPxqNLgqFQt2ibrFYrDEMo6Zi4ZObnf/Zn6Zp7pdeemnj7t27V+fzef9bb721oPxvhmG49u/ff0Mlx/mvf6ZpOvbu3XsjxuWcw+7du//tarvJZDJ45MiRRRhXkiR/S0vLEoyrKErNwYMHl2HcYrHofvvtt1E1KpVKzn379q3Anuc9e/Zc9fqOjo7Wt7a2LsC46XS69qOPPlqMcfP5vPf999//DMZVVdXzzjvvXF/p7Sv6jOvz+cyvf/3r72zatKmtrq5Ou/feewfK/9bS0nJtMpkMFAoF4S71bW1t19XX16NWYbz++usLJUmqzWazlT9LXebNN9+8NpvNBlOplE/U9fv9hf7+/vpcLifcbDaRSEyLx+PX5PP5gKibTCad0Wi08ciRI9NE3Z6enoZUKjXz0KFDi0XdDz/88LpEIlE/NDQkfJ4PHz48P5FI+HO5XMW70IrFovuVV15ZPzg42JjNZmv27NmzoFLXMAznzp07by8UCt6enp5ZZ86cmVKpa5qm429/+9utmUxmUjgcnvz6668vqdRljDlef/311fF4fJosy/5du3YtNwyj4sf1W2+9tSoajc5ctmzZ0CuvvLKyEqeqq8qZTMb185///OZ33323/tixY7NF/bGxMfmNN96o+AR9ksmTJyc/+OCDxgMHDkwXdVVVzf32t79d1NLSMlPU/eCDD+Y88cQT/yuRSAh3az558uTCjz76KGAYhvDPSTz//PMr9+7dO3NoaEjYXbx4sdTV1WUODAwIdR8vFAqOp556as2Pf/zjFUNDQ0IdplVVdX7/+99f/cILL0x/9913Kw6fw+Fgb7zxRm1nZ+fy1tbWKTt27LjZsqyKQuBwOPiePXtqwuHwjIGBgZr33ntvRqXjut1ufvDgQWdfX9/syZMnD7z88surVFWt6AnH6XTyw4cPs+7u7jlutzv/+9//fm2xWKz4yaqtra3U2tq64Je//OXKZ5555tZUKvWp97eq4HLOHTU1NXzatGmqZVnCrdLnzZsnOxwOCzN2OBwOFAoF71133SX8GzLLly8vrlixYkSWZa+oe/jw4c9ec8019eFwuOIHRZkZM2ac8ng8+T/84Q83iLqKotTNnj1bb29v/6yo63a7rZ6enml33HHHBRHPsiyHx+Mxpk6dGjcMQ+gdRm1tLXv44YdPe73eWaVSqeIHsdfrZbNnz85xzt1z5szRZs2aVfGYHo+Hz5kzJ+fz+djq1avzIvMFALj22msll8vFu7u7F82dO/dibW1txVeL5s2bJzmdTmhqakpf7h9dcbbmzZuX6+zsrO3q6roxGAzOjsfjn/pOsKrgTpkyxbz77rv7a2pqGleuXBkV9fP5/PyGhobGoaEh4VevwcHB5bfddpsjn88Lv909dOhQ08yZM/3r1q0bE3V/+tOf7nn88cd3LVu2bFjU7e3tnTxp0qTGL3zhC0Oi7je/+c2OxsZG95o1a/pE3Uwm47r99tuH5s2bp4p49fX17Ctf+Upk1apVbPr06UnRcWfMmOFct27d8Pr16/srdXRdd8+ZM2fO0NCQV1GU4Pz58/sr/cWVUqnkmjlz5rUdHR3z+/v7rw0Gg/MZYxW9Wpum6ZwyZcq8s2fPLohGo9ffdtttHtM0K3IZY45gMHhdf3//tW+//fYta9eujXk8nlIlLgCA1+u9bs6cOXU7d+58+Rvf+MaeefPmFT/NcXBO64UJwm7QyimCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAhFFyCsCEUXIKwIRRcgrAh/xdGVFaQoPVOigAAAABJRU5ErkJggg=="
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -431,14 +440,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"As the first move of the game, it's best to place the piece in the center of the board to maximize possible winning directions.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -459,8 +468,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -471,14 +482,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"As the game has just started, I will place my piece near Alice's to prevent her from forming a line and to start building my own line.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        8\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -499,8 +510,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -511,14 +524,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"My opponent placed his piece next to mine. I should block him and also try to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        6\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -539,8 +552,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -551,14 +566,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"I see Alice is trying to create a line horizontally. I need to block her next move otherwise she will have a chance to win. I will place my piece at [7, 5].\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        5\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -579,8 +594,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -591,14 +608,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [7, 9] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        9\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -619,8 +636,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -631,14 +650,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"Alice is trying to create a line horizontally again. I need to block her next move otherwise she will have a chance to win. I will place my piece at [7, 10].\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        10\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -659,8 +678,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAmkUlEQVR4nO2deXBU15m3f7cXdUvd2tnEHrBYBY5NzGKWfHaCwcHGYRBJGLCdZLJ4xi5PkkrNVFLJZGbK+eIJM6nJJIHKVBwnBmMw2A6LwRBjGRsQQgiQhGjQ0pJarW71vu/33vP9YZovHmPoc8QyF79PlcqF1c89R/ftXy/33nNfiTEGgiC0he52T4AgCH4ouAShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUG4g3vhwoWKrVu3LhoaGiq5GRMiCOL6SLwXYKxfv355ZWWlPHbs2OSPfvSjpkwmY2CM6W/S/AjiE4PBYJANBoNS0GN5Ny7LsmHGjBler9dbnslkdPv27fvM2bNnn6urq6vh3Zaqqjh//jzmzp3Lq4Ixhra2Ntx9993cLgC0trYKufF4HB6PB1OnTuV2U6kUBgYGMG3aNG43k8nAbrdj5syZ3K4sy7h06RJmz57N7aqqio6ODsyZM4fbZYyhvb39ltc3Go3C7/djypQp3G4ikYDL5UJtbS23m06n0dfXhxkzZnC7uVyOFRcX/99169ZtL0hgjHH9vPrqq5O/8Y1vfKm5uXkkYwyZTMa4bdu2NiZALpdjL730kojKVFVlf/jDH4TdF198Uch1uVzs4MGDQq7f72d79uwRcqPRKNu1a5eQm0wm2fbt24XcbDbLtm7dKuQqisL++Mc/CrnDqdHAwAA7fPiwkOvxeNj+/fuF3FAoxF5//XUhNx6Pyzt37nyKFZhD7nfcdevW9a1bt66P1yMI4sZBR5UJQoPc1uBKknQ7hycIzXJbg8toSSFBCEEflQlCg1BwCUKDUHAJQoNQcAlCg1BwCUKDUHAJQoPQeVyC0CDcwT127NiEH/3oRxtuxOB0HpcgxOC+VnnSpElum82mqqoqqaqKCxcujHc4HKZz585xD66qKhwOB0RcxpiwCwADAwNCrt/vh91uF3IjkQh6e3uF3GQyib6+PiE3m82iv79fyFUURdi9XTXyeDzC+yoUCgnXNx6PC9c3nU4jk8kU/Hju4EqSZMxkMmWRSMRcWlqaVlVVYoxBVVXeTUFV1Ss/IoiO+5dji4wpOu5wXFVVNecO5+/N+5+k/cwDd3B1Op35W9/6VmcgECiprKxM1dXVDdhstsy9997LuynIsgybzQYRl11e6ynqtrW1CblutxuKogi5gUAA0WhUyI3FYhgaGhJyU6kUent7hdxcLodLly4Jufm1vLe6Rk6nE3q9Xsj1er1Ip9NCbjgcRiAQEHITiQScTmfBj+cO7tixY0OrV69u4PUIgrhx0OkggtAgFFyC0CAUXILQIBRcgtAgdOUUQWgQugMGQWgQ+qhMEBqEgksQGoSCSxAahDu4oVDI5PV6y2VZpiNLBHGb4L7k8Xe/+12NzWa7b9myZee++tWvdt2MSREEcW24g/vEE08MPPPMM7PLysqS+aZfJ0+etOr1/A37VFXFyZMnYTQauV3GGJqammAymbhdAGhqaoLZbOb2wuEw+vv7EQ6Hud14PI5Lly4hmUxyu6lUCm1tbZBlmdvNZrM4ffq00Ok3RVHQ2NgIg4H7qXKlvkVFRdxuvr4iNQoEAnC5XAgEAtxuNBpFT08PYrEYt5tMJtHR0cG1PC9PJpPhqi13NX7605/OLy4unjxr1qxGk8mkPvTQQ63xeDzx2GOP8W4KiqIglUpBxGWMIR6PC7nAB6ttRNyhoSG0t7dj+fLl3G4wGERFRQVWrVrF7cZiMZhMJqE5p9NpABByZVlGJpMRclVVRTKZvOX1HRwcxKVLl/Dggw9yuz6fDy0tLVi5ciW3G4lEYLFYsHr1am43mUziz3/+c8GP5w7uv/zLv5zLZrOXLBZLDADMZnPOaDSy4uJi3k1BlmUUFRVBxGWMDcs1Go1CrtlsFh53OO5w9hUA4b83l8sJj6uq6ieqRplMZlj7Sqcr/JATd3ArKytTAFK8HkEQNw46HUQQGoSCSxAahIJLEBqEgksQGoSCSxAahIJLEBqEgksQGoSCSxAahIJLEBqEO7jHjx8fsX///nl+v5//6m+CIG4I3Jc82u32zAsvvDDtS1/6Uvrv/u7vOmRZ1quqKrRqRVEUKIoi5DLGhF3gg2tDb/Wch+PKsjyscUX/3uGMq6rqsOo7nDlrsb4892DjDu7ChQtTzc3N3mQyaby8rG9eY2NjKe92gP+/7EtkSSDwwdI8kSWBAISXE0YiETgcDgSDQW43kUigs7MT8Xic202n02hvb0c2m+V2c7kcWlpahJtRnTx5kusC+DyMMZw8eVJoSSAgXt9QKASXywWv18vtxmIx2O12RCIRbjeVSuHChQtIpfgv5eetK/ce3bdv36f0en3lokWLbCaTSV2zZs0pWZZjGzbwt8yVZRk6nQ6PP/44t5t/RRZ1FUURct1uN1pbW4WWfQUCARw/flxo2VcsFkNlZSXq6+u53VQqheLiYqxfv57bzeVy0Ov12LhxI7ebf6G41TVyOp2w2WxCSy+9Xi+am5uFll6Gw2E0NDRgzZo13G4ikcCbb75Z8OO5g/u9732vC8ANufMF3VeZIMSg+yoThAah00EEoUEouAShQah3EEFoEPqOSxAahD4qE4QGoeAShAah4BKEBqHgEoQGoaPKBKFBuIMbiUR03/nOdx51Op3Vwx2cjioThBjc1yrv2bPnrldeeWXiN77xjdaamppgT0/PaLfbbbx48SL34Kqqwu12Q8RljAm7wAc9gERcv98Pp9Mp5EYiEQwODgq5iUQCLpdLyM1kMsL7SlEUzdXI4/EI1ygYDAq7sVhMuL6pVIprhRB3cMPh8Pi77rprZG9vb8306dMHwuGwJZFI6EWWuamqing8LrREDvjgySzqio4bDocRi8WE3Gg0KuymUinhOWezWWFXURRhlzE2rBqJusOp0XDcRCIhvK/S6TTXskvu4D777LPv3HfffR0zZsyIG41GNm/ePLvdbk/ff//9vJuCLMvo7e2FiMsYQ1dXl7Db2dkp5LrdbhgMBiE3EAggm80KubFYDKFQSMhNpVJwu91Cbi6XQ39/v5Crqip6enpueY2cTidKSkqEXK/XC8aYkBsOhxGPx4XcRCIBv99f8OOFVjgvWrTII+IRBHFjoNNBBKFB6HQQQWgQWmRAEBqEPioThAah4BKEBqHgEoQGoeAShAaho8oEoUHoqDJBaBDu4Pb399ccPHhwYTweL74ZEyII4vpwX/L43nvvzW5ubjYvWrSIAcDlpl9SLpfjHjzfJEnEzbeoEHHzY4u4+SZY5F6ffNMv0Rqpqqqpv3e47k1t+jV+/Pizx44du/eFF16Y9cwzz5y73PTLKvJ9Nd9QSqQpFGMMTU1NKCoq4nYB4NSpUzCZTNxeOByGw+FAKBTiduPxODo7O5FIJLjddDqNtrY2oSdFLpfD6dOnuT3ggxc40cZs+RqJNmYTrW8wGITL5eK6aD9PNBqF3W5HNBrldpPJJC5cuIB0Os3tZjIZrsdzJ+bcuXMVFoulfNGiRR03oumXJElCDaXyTb9EXVmWhdzb2fSroqJCuOmX2Wy+LU2/GGO3vEbU9OsqfPe73+0B0MPrEQRx46DzuAShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUEouAShQbiD++677478x3/8x9VvvPHGqJsxIYIgrg/3tcr/9V//tVCSpOJ77rlHUhRF6u7uHuNyuYpsNhv34IqiwOVyQcRljAm7wAeLBURcn88Hh8Mh5IbDYQwMDAi5iUTiysXzvGQyGQwODgq5siwL72dVVYdVX9EaDQ0NCdcoGAwK1ygWiwnX6KY3/Zo4cWIGQPr8+fOz6+vrvbFYrDiVSukikQjvpqCqKpLJJERcxhhSqZSQC0DYjcViwnMejptMJoXnnM1mhV1FUahGBRKPx4XdW9H0q/H06dOza2trbQaDgd1zzz293d3d6YULF/JuCrIso6enByJuvimUqHvx4kUh1+12Q6/XC7mBQADpdFrIzXeQE3FTqRQGBweF3Fwuh97eXiFXVVV0dXXd8ho5nU4UFxcLuV6vF4qiCLn5Tn8ibiKRgNfrLfjx3MGdMmVKbMqUKSd5PYIgbhx0VJkgNAgFlyA0CAWXIDQIBZcgNIhQR/obxSehkwFjDLlcDi6XC5cuXcLFixcxffp0TJgwAcXFxdfcB3nX6/Wiv78f/f39sNvtqKmpgdlsLth1Op1wOBzcrsvlgtPpRFdXFy5dusQ9Z4/Hg4GBAaFxXS4Xuru7cenSJUycOPG67ieN2xrcO7mTQf6+z++++y42b96MpqYm+Hw+KIqCTZs2Yfbs2fj617+O+vr6j4Qh754+fRqdnZ0YNWoURo8ejc997nPo6+vDiRMnMHLkSCxduvSqrizLaGlp+ZC7YsWKK+6oUaOwZMmSq4Ywl8tdmfOpU6euzHnz5s2oq6vD3/zN3+Cv/uqvCh531apV1x037zY0NGDLli0fGve3v/3tNcf9xMIYG9ZPJpMxbtu2rY0JkMvl2EsvvSSiMlVV2R/+8Adh98UXXxRyXS4XO3jw4HUfl0wm2T/90z+x8vJyBuCqPyaTiW3cuJENDQ19xH311VfZsWPHWDqdZqqqfmjuuVyOdXR0sD/+8Y/M5/N96PeFuOfPn/9Y9yc/+QkrKyu75pw3bNjAPB7PR9ydO3cWNK7f77/ye1VVWSKRYD/+8Y+vO+7jjz/OvF7vh7Z9NQYGBtjhw4evW6Or4fF42P79+4XcUCjEXn/9dSE3Ho/LO3fufIoVmLvb+o57p5J/p3j++eeveRlbJpPB9u3bYTAYsHnzZhQXF0NRFBw+fBizZs3CrFmzPvLuIkkSDAYDZs6ciaqqKhw8eBDr1q2D2WyGoig4dOjQdd1Zs2ahurr6I+6WLVvw/PPPX/Pm3JlMBq+88gqKioqwefNmmM1mqKqKt956C3V1dZg5c+Y1x62qqsKBAweujMsYw5YtW/Dzn//8uuO+/PLLMJlM+NWvfgWz2XytEtzx0MGpGwxjDJcuXcKmTZsKuvZUVVXs3LkTBw4cAGMMvb29V57k1/pIKEkSRo8ejTlz5uDkyZNgjMFut6OoqIjLbWpq+tCcC7mjvqqq2LFjB9566y0wxtDd3Q2z2YwZM2Zcd9wxY8Zg9uzZOHXqFBhjsNlsXONu27YNhw8fvqO/ZhUCd3BjsZgxFAqVplIpsb4SnwDeeOMNuN3ugh+fSqXw0ksvQVEUtLa2Yv78+QV9j5MkCbNmzYLD4YAsy8JuLpfD7t274fF4uOa8detWZLNZtLa2YsGCBdDprv90kiQJdXV16O/vRzabxWuvvcZ1qV86nca2bdsgy3LBzp0Id3Db2tpGP/XUU19++eWX627GhLSOqqo4fvw49zvChQsX4PV6kUwmUV1dXbBnNBpRVlYGv9+PdDqNqqoqLtdqtSIQCOD06dPccz579iyCwSByuRzKy8u5xi0tLUUoFMKxY8e4xz1z5ozwwoU7Be7vuIsXL3b+/Oc/Dy9durQnk8noDh48eM+pU6esIg20VFVFc3Mziov5O3YyxtDc3AyLxcLtAsDp06dhtVq5vXA4jL6+PsTj8av+PpfLoa+vj3u7Ho8HO3bsgMlkKuidK48kSTAajdi1axeMRiPXEVdJkqDX67Fr1y7hZYq7d++GyWTiPtIrSRJef/11dHZ2co/r8Xjw+uuvf+yLVCAQwODgoFC4o9Eouru7kUqluN1kMonz589DURRuN5PJcDX+4g5uZ2en1Ww2y9OnT48CwLJly2yBQCC5YsUK3k1BURREIhGIuIwxBINBIRcA/H6/kOvxeNDe3o7Pf/7zV/29oih44YUXuINQXV2NVatWobGxEaqqFhxedvnc58qVK3HyJN/aD3b5tNPDDz+MAwcOoLu7m8svLS3FihUrrnxf5R37oYcews6dO9Hf38/lVldXY/ny5RgxYsRVfz84OIjOzk488MADXNsFPnhelJWVCT03IpEIDAaDkJtMJtHQ0FDw47mDO27cuOSvfvWrt/L/tlqtabPZrJaWlvJuCrIsw2w2Q8RljN0WNx6Po7i4+GNdxhgWLVqEI0eOcG13+vTpmDJlCtra2hAOhwv+yCvLMqLRKCZPnoxz584hEomgoqKiYDcej2PixImYN28eDh06xDXnOXPmYMKECTh79izi8XjBH5cVRUE8Hse4ceOwZMkSvP/++1zB//SnP41x48Z9bAtOq9V6zRpdi1QqJewqiiLs6nQ6rk9a3N9xLRaLOmrUKP4GoFfhTj2RvnbtWowcObLgxxcVFWHjxo0wGAyoq6tDS0tLwU/k7u5u1NTUwGg0oq6uDs3NzQW5jDF0dXWhpqYGRUVFqK+v5/puXVRUhA0bNsBsNnOPe/HixStXUq1duxaVlZVc465fv1645+6dwm09HXQnHtLPHzV95plnCmrYLUkSHn30UTz22GOQJAm1tbUIh8Ow2+3X3D/5rwqnTp3C/fffD0mSMG3atILdUCiE5uZmLFq0CJIkYc6cOXj66acLCoQkSVi9ejUeffRRSJKE6dOnIxgMore3t6BxW1pasHDhQkiShLlz53KN+8UvfhGrVq26Y1/0C+W2BvdO3fkGgwHf/e538dRTT13zwJter8cXvvAF/OIXv7jy8cpgMODhhx/GyZMnce7cOciy/KEwsMsNvfv7+7F3714sX74cJSUlV3UVRbmq29fXd1X3+9//Pv72b//2mhc35Mf4xS9+ceXAYP7/nThxAq2trR87bm9vL/bu3YuHHnroyrh6vR7f//738e1vf/u64z7yyCP4j//4D+EDkncSdK3yTcJqtWLTpk1YunQptmzZgtbWVkQiETDGYLVaMWXKFDz55JP42te+9pHvhhaLBfX19Th+/Dh27NiBiRMnoqamBnq9Hj6fD319fTCZTFizZg3Kyso+9AL4l+4rr7zC5VqtVvzbv/0bFi9ejN/85jdob2//0JynTp2Kr371q3jyySdRXl7+Efda4/b29qK4uPgj40qShNLSUmzatAmLFy/Gli1buMb9pEKXPN4kJEmC2WzGunXrsGrVKtjtdpw/fx4dHR1YvXo1amtrUV5eftUDEpIkwWQy4YEHHkAymYTT6URvby9aWlrwyCOPYMWKFSgrKyvYdTgcaGxsxGOPPXZdNz/nL3zhC7Db7bDb7Xjvvffwla985bpzNpvNePDBB6+M63K50NDQgLVr12LlypXXHffLX/4yHnnkEdjtdvT19eHIkSPYsGHDNcf9pELBvclIkgSLxYI5c+Zg7NixsFgsmD9/Ppc7ffp0jB07FtFoFHPmzOF2J06cCK/Xy+VarVbMnTsXM2fORDweF5pzbW0t+vv7hcadM2cOgsFgweN+0qCXMILQIBRcgtAgFFyC0CAUXILQINzBdTgcZW+//fbdnZ2dJcMdnA7rE4QY3EeVf/KTn9x39OjRaU899ZTuH/7hH86qqirlT7DzoqoqRN38LTxE3Lx/q8cdjnu79tVwxtXinIdbX1VVhzXnQuEOrqIorLq6OmAwGEyZTEb3xhtvzD9+/Hip6GQbGxu5PeCDnXTixAnhc3snTpyAXq/n9sLhMBwOB3w+H7cbj8fR1dUl3BSqvb1daLlZNpvFmTNnhBafK4py5S4ZvKiqihMnTgh9smKMobGxUahGwWAQbrcbQ0ND3G40GoXdbkcwGOR2k8kkLly4gEQiwe1mMhmu5zJ3cJcuXRqMx+NjKisrvSaTSV27dm2TqqqxDRs28G4KsixDr9fj8ccf53bzT6QnnnhCyFVVVch1u91obW3FypUrud1AIIDjx49j9erV3G4sFkNVVRXq6+u53VQqBYvFgvXr13O7uVwORqMRGzdu5HbzyxNFa8QYE3LzrS6XL1/O7Xq9XjQ3N2PVqlXcbjgcRkNDA9asWcPtJhIJvPnmmwU/nju43/zmN89985vfPMfrEQRx46CjygShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUEouAShQSi4BKFBKLgEoUHoLo8EoUG4g7tv376pzz777Oq2trbC72L9MdzJd3kkiJsJ97XK27dvr3U6nezVV1+9d+7cuUccDscIn89ntNvt3IMrigKv1wsRlzEGn88n5AIQdn0+H4aGhoTccDgMj8cj5CYSCeF9lclkhF1ZljVXo6GhIeEaBYNBYTcWiwnXN5VKIZfLFfx47uA+++yzp7ds2fKg0Wj05XI5yeVyVQYCAcPAwADvpqCqKoLBIETc/J38RVwAwm4wGITP5xNyo9GosJtMJuH3+4XcbDaLQCAg5CqKMqwaiY4LiNfI5/MJ7+dwOCy8n+PxuLCbTqe5ll1yB3fEiBH6Bx54wLV8+fJzRqORLVy4sGtgYCD12c9+lndTkGUZTqcTIi5jDP39/cJub2+vkOt2u1FcXCzkBgIBMMaE3FgshkQiIeSmUin4/X4hN5fLweVyCbmqqsLhcNzyGjmdTpSXlwu5Xq8XBoNByA2Hw8hms0LuTV/WV1tb66mtrS28dTlBEDccOh1EEBqEgksQGoTO4xKEBqH+uAShQegdlyA0CL3jEoQGoYNTBKFBKLgEoUHoOy5BaJCCgpvJZPQNDQ13ezwey9tvvz2jqalp9I0YnL7jEoQYBV3yqCgKfvazn9319NNPW3bu3Dkqk8kU7dq1a5dOp2OqquoYY1AUhXvwfIMkERf4IPi32h3OnIfrDmfOt+Pvzbd6uR010mp9C6Wg4JaUlChTp051AzDPnz/f39/fPw4ALjf9um+4Tb9Em0INp+mXaEOpSCSC/v5++P1+bjff9CsajXK7w2n6lcvl0NLSMqymXyLcrhoNt+lXb28vQqEQt5tMJmGz2f73NP1Kp9NGi8Uy9fz585He3l7zmDFjXDqdjplMJkZNvwrndjb9KikpuS1NvyRJuuU1oqZflzGbzbl///d/38o9G4Igbgp0VJkgNAhdOUUQGoQuwCAIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1CwSUIDULBJQgNUlBwY7GY8Qc/+MG61tbWsX//93+/9tixYyNv9sQIgvh4CrpWubS0NDcwMOBljJlramr8Fy9etC5ZssQHAP39/SO9Xq+xp6eHe/B80y8RlzEm7AIf9JcRcX0+H9xut5AbDocxNDQk5CYSCXg8HiE33/RLxM03/RJxVVW9LTUaGhoSrlF+ZZGIG4vFhOt705p+mUwmYzqdNldUVBiz2awZAHK5nDQ0NFQeCoUMg4OD3JNVVRWhUAgiLmNM2AUg7AaDQQQCASE3Go0Ku8lkEsFgUMjNZrPCbr7pl5Zq5PP54Pf7hdxwOCxco3g8LryvbkrTr0wmY1i+fLk6ODhoraioUKuqqiwAYDQa2YIFC7odDkdq2bJl3JOVZRkDAwMQcRlj6OvrE3btdruQ63a7YTabhdxAIABVVYXcWCyGeDwu5KZSKfh8PiE3l8thcHBQyFVVFf39/be8Rk6nE2VlZUKu1+uFXq8XcsPhMDKZjJB7U5b1mUwm+Stf+co73LMhCOKmQEeVCUKDUHAJQoPQQnqC0CC0kJ4gNAh9VCYIDULBJQgNQsElCA1CwSUIDULBJQgNUnBwE4mEWZZlXTweNyeTSf6+EARB3DAKCm4kEimqr6/feOLEidof/OAH/+fZZ5996GZPjCCIj6ega5XLy8uzEydOtI0YMSJbX19v+/Wvfz0LALLZrO7w4cNzz5w5Y7FardyDq6qKM2fOoKysjNtljOHMmTOoqKjgdgHg7NmzqKys5PZCoRB6e3uRyWS43VgsBpvNJnT+OpVK4ezZszAajdxuNptFS0sLSkpKuF1FUdDS0oLS0lJuN1+j8vJybhcQr1EgEIDT6UQymeR2I5EIurq6hBqkJRIJtLe3CzU5y2QyXM3CCgpuNpvVx+Pxie+++27p73//+7nPPPPMmwBQVFSkzp8/v8flcqWWLl3KPVlFUeDxeCDiMsbgcrmEXOCDFSQirtfrhcViEXJDoRB0Op2QG4/HkU6nhdxMJoNIJCLkyrIMn88n5KqqCrfbLVyjwcFBIdflcqGyslLI9fv9KCoqEnKj0SgURRFyk8kk3n///YIfX1Bw9Xq9+tOf/vSooihsxYoVHWaz+cpLQ0VFRcJisahVVVXck5VlGRaLBSIuY+y2uJlMBqWlpcLjWq1WIddoNAq7qVRK+O/N5XLCrqqqw6pRSUmJkJtMJoX3lSzLwvXV6XTC45pMJq6WooUGl02ePNnFPRuCIG4KdDqIIDQIBZcgNAgFlyA0CAWXIDQIBZcgNAgFlyA0CAWXIDQIBZcgNAgFlyA0SEHBzWazuqamphl+v9/a1NQ0vbW1dcSNGJzu8kgQYhR0yWMul5N++MMfzv3nf/5n5eLFixX79++/Z8+ePTsAgDEmXf4v9+CMsSs/Iq7ouP9zGyLerR5Xi3/vcOr7l9sQdW5HfYf79xZKQcG1WCzKtGnTnJWVlXJtbW1AkqTJAJDJZHSvvfbagvfff79UZBmUqqpobGwULs7x48e5vb90Rd7xw+EwBgYG4PV6ud14PI6uri6EQiFuN51Oo729nWvpV55sNoszZ84gm81yu4qioKmpCaqqcrv5+oownBrlO+65XPyX10ejUfT29iIQCHC7yWQSNpsNsViM281kMlzLAQsKbjqdNhoMhtq9e/eO0ul0yl133RUAAJPJpNbX159kjMU2bNjAPVlZlmEwGPD4449zu4wxSJKEJ598UsgFIOS63W60trZi5cqV3G4gEMDx48exevVqbjcWi+HQoUOor6/ndlOpFKxWK9avX8/t5nI5FBUVYePGjdyuqqrQ6/V44oknuN3h1MjpdMJms2H58uXcrtfrRXNzM1atWsXthsNhNDQ0YM2aNdzuTWn6ZTabc7/85S9fyu9MvV5PdzIniNtIwf1xdTodhZUg/pdAvYMIQoNQ7yCC0CB0AQZBaBAKLkFoEPqOSxAahL7jEoQGoY/KBKFBKLgEoUEouAShQQoKbjweNz733HNf9Hg8VYcPH57ym9/8Zt7NnhhBEB9PQZc8mkwmuaOjI3L+/PlRv/zlL+cqipJ5+umnWwDA7XZXhEIhyel0ci89URQFwWBQcjqd3EepGGO4PK7QES5R1+v1wu/3C7nhcBiBQEDITSQSwvsqk8kIu7IsD6tGoi4gXiO32y1co0AgIOzGYjHh+qZSKZVnhV1BwTUajcxqteKdd94ZV11dPbazs1MfDAZ1paWlrLu7e/Tg4ODLNpvtDO9kVVXVuVyue20222lelzEmuVyuz9hstmZeFwBcLtd9Im4wGCwbGBgYa7PZLvK6sVjMMjAw8CmbzXae102lUman0zndZrO18rrZbNY4ODhYZ7PZzvK6iqLoXS7Xp202WwuvO8waCbt+v7/C4XCMstlsnbxuOBy2DgwMTLTZbBd43UQiUex0Ou+y2WztvG46nS7KZrNFBQt/udj5435SqZTh97///co9e/YsiEajxX/605+m5H+XzWb1b7755pxCtvM/f2RZlvbt23e3iMsYw549ez59q12fz2c9duzYXSJuOBwubmhomC7ixuPxosOHD88ScdPptOHAgQNCNcrlcrr9+/fPFd3Pe/fuveX1HRoaKmtsbJwi4gYCgZL33ntvmogbjUZNb7/99kwRN5lMGt96663ZhT6+oO+4ZrNZ/trXvvbW6tWrm0pLS1OPPfaYPf+7hoaGCT6fz5JIJLi71Dc1NX2qrKxM6CqM3bt3Tw2HwyWhUKjwV6nLvPHGGxNCoZDV7/ebed3i4uJET09PWSQS4W426/V6R7rd7nHRaNTC6/p8Pp3T6aw8duzYSF63o6Ojwu/3j3nnnXem8brvvvvup7xeb1lfXx/3fj569Ohkr9dbHIlECl6Flk6nDdu2bVvW29tbGQqFivbu3TulUDebzeq2b9++JJFImDo6OmrOnTtXXagry7K0Y8eO+4PBYLnD4ajavXv39EJdVVWl3bt3L3C73SNjsVjxzp0767LZbMHP6z/96U/znE7nmFmzZvVt27bt3kKcYR1VDgaD+k2bNt136NChshMnTozl9T0eT+y1114reAf9JVVVVb4jR45UHjx4cBSvm0wmI5s3b76roaFhDK975MiR8d/73vc+5/V6ubs1nz59eup7771nyWaz3LeT+PWvf33vvn37xvT19XG706ZNC7e1tcl2u52r+3gikZB++MMfLvzXf/3XuX19fVwdppPJpO7HP/7xgt/97nejDh06VHD4JElSX3vttZLW1ta6xsbG6q1bt96nKEpBIZAkie3du7fI4XCMttvtRX/+859HFzquwWBghw8f1nV1dY2tqqqyv/jii/OSyWRBLzg6nY4dPXpUbW9vH28wGKK//e1vF6XT6YJfrJqamnKNjY1T/vM///Pe55577n6/33/dv3dYwWWMSUVFRWzkyJFJRVG4W6VPmjQpJkmSIjK2w+GwJBIJ0+c//3nue8jU1dWl586dOxiLxUy87tGjRz8zbty4MofDUfCTIs/o0aPPGI3G6H//93/P4XXj8Xjp2LFjM83NzZ/hdQ0Gg9LR0THys5/97CUeT1EUyWg0ZkeMGOHOZrNcnzBKSkrUb33rW2dNJlNNLpcr+ElsMpnUsWPHRhhjhvHjx6dqamoKHtNoNLLx48dHzGazumDBgijPfAFgwoQJYb1ez9rb2++aOHFid0lJScFHiyZNmhTW6XSora0NXO4fXXC2Jk2aFGltbS1pa2u722q1jnW73df9JDis4FZXV8srVqzoKSoqqrz33nudvH40Gp1cUVFR2dfXx/3u1dvbW7d48WIpGo1yf9x95513aseMGVO8dOlSD6/7/PPP7/3Od76zc9asWf28bmdnZ1V5eXnlww8/3Mfrfvvb326prKw0LFy4sIvXDQaD+iVLlvRNmjQpyeOVlZWpf/3Xfz0wb948ddSoUT7ecUePHq1bunRp/7Jly3oKdTKZjGH8+PHj+/r6TPF43Dp58uSeQu+4ksvl9GPGjJnQ0tIyuaenZ4LVap2sqmpB79ayLOuqq6snnT9/forT6Zy9ePFioyzLBbmqqkpWq/VTPT09Ew4cODB/0aJFLqPRmCvEBQCTyfSp8ePHl27fvv3Fr3/963snTZqUvp4jMUbXCxOE1qArpwhCg1BwCUKDUHAJQoNQcAlCg1BwCUKDUHAJQoNQcAlCg1BwCUKDUHAJQoP8PzS/I+ovoNUlAAAAAElFTkSuQmCC",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -671,14 +692,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [6, 7] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        6,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -699,8 +720,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAoS0lEQVR4nO2deZRU5Z33v7eWruqu6o1ma/YAzdqgQkRahLyaIBgUJTYmDKhJJoszepwkJ2fmJCeZzMwx7zhhJmcySeBkToyJIoqihEUQorQL0DRNA71R0EtVL9VVXfu+33uf9w8t3hhZ6nmaJRd/n3P6eLDu5z5P3ae+tdz7PPcnMcZAEIS20N3oDhAEwQ8FlyA0CAWXIDQIBZcgNAgFlyA0CAWXIDQIBZcgNAh3cM+ePVvx4osv1g0PD5dciw4RBHFlJN4JGBs2bFhZWVkpT5gwIfmjH/2oKZPJGBhj+mvUP4L41GAwGGSDwaAUtC3vzmVZNsyZM8fr9XrLM5mMbu/evZ89ffr0M7W1tdW8+1JVFR0dHVi4cCGvCsYY2tracMstt3C7ANDa2irkxuNxeDwezJgxg9tNpVIYHBzErFmzuN1MJgO73Y65c+dyu7Is4/z585g/fz63q6oqOjs7sWDBAm6XMYb29vbrPr7RaBR+vx/Tp0/ndhOJBFwuF2pqarjddDqNvr4+zJkzh9vN5XKsuLj4/65fv357QQJjjOvv1VdfnfaNb3zjkebm5jGMMWQyGeO2bdvamAC5XI698MILIipTVZX9/ve/F3aff/55IdflcrEDBw4IuX6/n+3evVvIjUaj7LXXXhNyk8kk2759u5CbzWbZiy++KOQqisL+8Ic/CLkjGaPBwUF26NAhIdfj8bB9+/YJuaFQiL3xxhtCbjwel3fs2PEEKzCH3J+469ev71u/fn0fr0cQxNWDzioThAa5ocGVJOlGNk8QmuWGBpfRkkKCEIK+KhOEBqHgEoQGoeAShAah4BKEBqHgEoQGoeAShAah67gEoUG4g3vkyJHJP/rRjzZejcbpOi5BiME9V3nq1Klum82mqqoqqaqKs2fPThoYGDCdOXOGu3FVVTEwMAARlzEm7ALA4OCgkOv3+2G324XcSCQCh8Mh5CaTSfT19Qm52WwW/f39Qq6iKMLujRojj8cjfKxCoZDw+MbjceHxTafTyGQyBW/PHVxJkoyZTKYsEomYS0tL06qqSowxqKrKuyuoqnrhTwTRdv+8bZE2RdsdiauqqubckTzfvP9pOs48cAdXp9OZv/Wtb3UFAoGSysrKVG1t7aDNZsssWrSId1eQZRk2mw0iLvtoraeo29bWJuS63W4oiiLkBgIBRKNRITcWi2F4eFjITaVScDgcQm4ul8P58+eF3Pxa3us9Rk6nE3q9Xsj1er1Ip9NCbjgcRiAQEHITiQScTmfB23MHd8KECaG1a9c28HoEQVw96HIQQWgQCi5BaBAKLkFoEAouQWgQmjlFEBqE7oBBEBqEvioThAah4BKEBqHgEoQG4Q5uKBQyeb3eclmW6cwSQdwguKc8/va3v6222Wy3r1ix4sxXv/rV7mvRKYIgLg93cB977LHBp556an5ZWVkyX/Tr+PHjVr2ev2Cfqqo4fvw4jEYjt8sYQ1NTE0wmE7cLAE1NTTCbzdxeOBxGf38/wuEwtxuPx3H+/Hkkk0luN5VKoa2tDbIsc7vZbBYnT54UuvymKAoaGxthMHC/VC6Mb1FREbebH1+RMQoEAnC5XAgEAtxuNBpFb28vYrEYt5tMJtHZ2cm1PC9PJpPhGlvu0fjpT3+6pLi4eNq8efMaTSaTeu+997bG4/HEgw8+yLsrKIqCVCoFEZcxhng8LuQCH662EXGHh4fR3t6OlStXcrvBYBAVFRVYs2YNtxuLxWAymYT6nE6nAUDIlWUZmUxGyFVVFclk8rqP79DQEM6fP4977rmH2/X5fGhpacHq1au53UgkAovFgrVr13K7yWQSf/rTnwrenju4//qv/3omm82et1gsMQAwm805o9HIiouLeXcFWZZRVFQEEZcxNiLXaDQKuWazWbjdkbgjOVYAhJ9vLpcTbldV1U/VGGUymREdK52u8FNO3MGtrKxMAUjxegRBXD3ochBBaBAKLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQCi5BaBAKLkFoEAouQWgQ7uAePXp09L59+xb7/X7+2d8EQVwVuKc82u32zHPPPTfrkUceSf/93/99pyzLelVVhVatKIoCRVGEXMaYsAt8ODf0evd5JK4syyNqV/T5jqRdVVVHNL4j6bMWx5fnHmzcwV26dGmqubnZm0wmjR8t61vc2NhYyrsf4P8v+xJZEgh8uDRPZEkgAOHlhJFIBAMDAwgGg9xuIpFAV1cX4vE4t5tOp9He3o5sNsvt5nI5tLS0CBejOn78ONcE+DyMMRw/flxoSSAgPr6hUAgulwter5fbjcVisNvtiEQi3G4qlcLZs2eRSvFP5ecdV+4junfv3s/o9frKuro6m8lkUtetW3dCluXYxo38JXNlWYZOp8Ojjz7K7ebfkUVdRVGEXLfbjdbWVqFlX4FAAEePHhVa9hWLxVBZWYn6+npuN5VKobi4GBs2bOB2c7kc9Ho9Nm3axO3m3yiu9xg5nU7YbDahpZderxfNzc1CSy/D4TAaGhqwbt06bjeRSODNN98seHvu4H7ve9/rBnBV7nxB91UmCDHovsoEoUHochBBaBAKLkFoEKodRBAahH7jEoQGoa/KBKFBKLgEoUEouAShQSi4BKFB6KwyQWgQ7uBGIhHdd77znQecTmfVSBuns8oEIQb3XOXdu3fPfPnll6d84xvfaK2urg729vaOc7vdxnPnznE3rqoq3G43RFzGmLALfFgDSMT1+/1wOp1CbiQSwdDQkJCbSCTgcrmE3EwmI3ysFEXR3Bh5PB7hMQoGg8JuLBYTHt9UKsW1Qog7uOFweNLMmTPHOByO6tmzZw+Gw2FLIpHQiyxzU1UV8XhcaIkc8OGLWdQVbTccDiMWiwm50WhU2E2lUsJ9zmazwq6iKMIuY2xEYyTqjmSMRuImEgnhY5VOp7mWXXIH9+mnnz58++23d86ZMyduNBrZ4sWL7Xa7PX3nnXfy7gqyLMPhcEDEZYyhu7tb2O3q6hJy3W43DAaDkBsIBJDNZoXcWCyGUCgk5KZSKbjdbiE3l8uhv79fyFVVFb29vdd9jJxOJ0pKSoRcr9cLxpiQGw6HEY/HhdxEIgG/31/w9kIrnOvq6jwiHkEQVwe6HEQQGoQuBxGEBqFFBgShQeirMkFoEAouQWgQCi5BaBAKLkFoEDqrTBAahM4qE4QG4Q5uf39/9YEDB5bG4/Hia9EhgiCuDPeUx/fff39+c3Ozua6ujgHAR0W/pFwux914vkiSiJsvUSHi5tsWcfNFsMi9MvmiX6JjpKqqpp7vSN1rWvRr0qRJp48cObLoueeem/fUU0+d+ajol1Xk92q+oJRIUSjGGJqamlBUVMTtAsCJEydgMpm4vXA4jIGBAYRCIW43Ho+jq6sLiUSC202n02hraxN6UeRyOZw8eZLbAz58gxMtzJYfI9HCbKLjGwwG4XK5uCbt54lGo7Db7YhGo9xuMpnE2bNnkU6nud1MJsO1PXdizpw5U2GxWMrr6uo6r0bRL0mShApK5Yt+ibqyLAu5N7LoV0VFhXDRL7PZfEOKfjHGrvsYUdGvi/Dd7363F0Avr0cQxNWDruMShAah4BKEBqHgEoQGoeAShAah4N7E5M/MptNp5HI55HI5mq12kyB0zynirxvGGCKRCE6ePAmPxwNJkpBKpfDqq6+irKwMixcvRnV1Nc0V1zAU3JsMxhjOnDmDjo4OLFmyBCtWrLgwwUVVVQwPD+Po0aOoqqr62GOEtqCvyjcR+dAODAzgkUcewaxZs1BUVASdTgedTgeDwYBJkybhS1/6EkwmE95++22ue/kSfz1QcG8iQqEQOjo6cN9998FkMl3yq7Ber0ddXR1kWUZPTw/97tUgFNybiJMnT6Kurq6gucE6nQ4rVqxAS0sLfepqEO7gvvvuu2P+6Z/+ae2uXbvGXosOEWIoigKPx4OpU6cWfNLJarVCp9MhlUpd494RVxvuMxP/8z//s1SSpOLbbrtNUhRF6unpGe9yuYpsNht344qiwOVyQcRljAm7wIeLBURcn8+HgYEBITccDmNwcFDITSQSFybPX4xUKgVFUbhONkmSBIvFgpaWFowde/H3YVmWhY+zqqojGl/RMRoeHhYeo2AwKDxGsVjssmN0Oa550a8pU6ZkAKQ7Ojrm19fXe2OxWHEqldJFIhHeXUFVVSSTSYi4jDGkUikhF4CwG4vFhPs8EjeZTF62z5lMBrIsc+9XURQkEolL7ldRFBqjAonH48Lu9Sj61Xjy5Mn5NTU1NoPBwG677TZHT09PeunSpby7gizL6O3thYibLwol6p47d07Idbvd0Ov1Qm4gEEA6nRZy8xXkLuWqqorBwUFks9mC1xnnq+k99NBDKCkpueg2uVwODodDqM+qqqK7u/u6j5HT6URxcbGQ6/V6oSiKkJuv9CfiJhIJeL3egrfnDu706dNj06dPP87rEdcWSZIwZcoUdHV1oba2tqDfucFgEEajEWaz+Tr0kLia0FnlmwRJkrBo0SKcOnUKyWTyitvLsoyGhgYsXboUOh29DLQGjdhNhMViwZ133ok9e/YgGo1e9PosYwyZTAaHDh3ChAkTMGnSpBvQU2Kk0Hy3mwhJkjBz5kwYDAbs2rULNTU1mDt3LsrKyiBJEhKJBBwOB86cOYPa2lrceuutNF9Zo1BwbzIkScK0adNQXV2Nc+fO4fDhw4hGoxgaGkJNTQ0mT56MdevWwWq1Umg1zA0N7qfhhcMYQy6Xg8vlwvnz53Hu3DnMnj0bkydPRnFx8WWPQd71er3o7+9Hf38/7HY7qqurYTabL+lKkgSTyYR58+Zh9OjRGBwchMfjwe23335F9y/77HQ60d3djfPnz3P32ePxYHBwsKA+/2W7LpcLPT09OH/+PKZMmXJF99PGDQ3uzTxHNn/f53fffRdbtmxBU1MTfD4fFEXB5s2bMX/+fHz9619HfX39J8KQd0+ePImuri6MHTsW48aNw+c//3n09fXh2LFjGDNmDJYvX35RV5ZltLS0fMxdvXr1BXfs2LG46667LhrCXC53oc8nTpy40OctW7agtrYWf/u3f4svfelLBbe7Zs2aK7abdxsaGrB169aPtfub3/zmsu1+amGMjegvk8kYt23b1sYEyOVy7IUXXhBRmaqq7Pe//72w+/zzzwu5LpeLHThw4IrbJZNJ9s///M+svLycAbjon8lkYps2bWLDw8OfcF999VV25MgRlk6nmaqqH+t7LpdjnZ2d7A9/+APz+Xwfe7wQt6Oj45LuT37yE1ZWVnbZPm/cuJF5PJ5PuDt27CioXb/ff+FxVVVZIpFgP/7xj6/Y7qOPPsq8Xu/H9n0xBgcH2aFDh644RhfD4/Gwffv2CbmhUIi98cYbQm48Hpd37NjxBCswd/Qb9xqQ/6R49tlnLzuNLZPJYPv27TAYDNiyZQuKi4uhKAoOHTqEefPmYd68eZ/4dJEkCQaDAXPnzsWoUaNw4MABrF+/HmazGYqi4ODBg1d0582bh6qqqk+4W7duxbPPPnvZm3NnMhm8/PLLKCoqwpYtW2A2m6GqKt566y3U1tZi7ty5l2131KhR2L9//4V2GWPYunUrfvazn12x3Zdeegkmkwm//OUvP/XXnuly0FWGMYbz589j8+bNBc09VVUVO3bswP79+8EYg8PhuPAiv9xXQkmSMG7cOCxYsADHjx8HYwx2ux1FRUVcblNT08f6XMgd9VVVxSuvvIK33noLjDH09PTAbDZjzpw5V2x3/PjxmD9/Pk6cOAHGGGw2G1e727Ztw6FDh27qn1mFwB3cWCxmDIVCpalUSqyuxKeAXbt2we12F7x9KpXCCy+8AEVR0NraiiVLlhT0O06SJMybNw8DAwOQZVnYzeVy2LlzJzweD1efX3zxRWSzWbS2tuKOO+4oaCKHJEmora1Ff38/stksXn/9da6pful0Gtu2bROal30zwR3ctra2cU888cSXX3rppdpr0SGto6oqjh49yv2JcPbsWXi9XiSTSVRVVRXsGY1GlJWVwe/3I51OY9SoUVyu1WpFIBDAyZMnuft8+vRpBINB5HI5lJeXc7VbWlqKUCiEI0eOcLd76tQp4YULNwvcv3GXLVvm/NnPfhZevnx5byaT0R04cOC2EydOWEUKaKmqiubmZhQX81fsZIyhubkZFouF2wU+XHRutVq5vXA4jL6+PsTj8Ys+nsvl0NfXx71fj8eDV155BSaTiWsKoiRJMBqNeO2112A0GrnOuEqSBL1ej9dee014meLOnTsve7eNy7X9xhtvoKuri7tdj8eDN95445JvUoFAAENDQ0Lhjkaj6OnpEVqjnEwm0dHRAUVRuN1MJsNV+Is7uF1dXVaz2SzPnj07CgArVqywBQKB5KpVq3h3BUVREIlEIOIyxhAMBoVcAPD7/UKux+NBe3s7vvCFL1z0cUVR8Nxzz3EHoaqqCmvWrEFjYyNUVS04vOyja5+rV6/G8eN8az/YR5ed7rvvPuzfvx89PT1cfmlpKVatWnXh9ypv2/feey927NiB/v5+LreqqgorV67E6NGjL/r40NAQurq6cPfdd3PtF/jwdVFWVib02ohEIjAYDEJuMplEQ0NDwdtzB3fixInJX/7yl2/l/221WtNms1ktLS3l3RVkWYbZbIaIyxi7IW48HkdxcfElXcYY6urq8M4773Dtd/bs2Zg+fTra2toQDocL/soryzKi0SimTZuGM2fOIBKJoKKiomA3Ho9jypQpWLx4MQ4ePMjV5wULFmDy5Mk4ffo04vF4wV+XFUVBPB7HxIkTcdddd+GDDz7gCv6tt96KiRMnXrIEp9VqvewYXY5UKiXsKooi7OZv6Ffw9rwNWCwWdezYsfwFQC/CzXoh/eGHH8aYMWMK3r6oqAibNm2CwWBAbW0tWlpaCn4h9/T0oLq6GkajEbW1tWhubi7IZYyhu7sb1dXVKCoqQn19Pddv66KiImzcuBFms5m73XPnzl2YSfXwww+jsrKSq90NGzYI19y9Wbihl4NuxlP6+bOmTz31VEG3kZEkCQ888AAefPBBSJKEmpoahMNh2O32yx6f/E+FEydO4M4774QkSZg1a1bBbigUQnNzM+rq6iBJEhYsWIAnn3yyoEBIkoS1a9figQcegCRJmD17NoLBIBwOR0HttrS0YOnSpZAkCQsXLuRq96GHHsKaNWtu2jf9Qrmhwb1ZD77BYMB3v/tdPPHEE5c98abX6/HFL34RP//5zy98vTIYDLjvvvtw/PhxnDlzBrIsfywM7KOC3v39/dizZw9Wrlx54e4Vf+kqinJRt6+v76Lu97//ffzd3/3dZSc35Nv4+c9/fuHEYP7/HTt2DK2trZds1+FwYM+ePbj33nsvtKvX6/H9738f3/72t6/Y7v3334//+q//Ej4heTNBc5WvEVarFZs3b8by5cuxdetWtLa2IhKJgDEGq9WK6dOn4/HHH8fXvva1T/w2tFgsqK+vx9GjR/HKK69gypQpqK6uhl6vh8/nQ19fH0wmE9atW3dhyd7F3JdffpnLtVqt+I//+A8sW7YMv/71r9He3v6xPs+YMQNf/epX8fjjj6O8vPwT7uXadTgcKC4u/kS7kiShtLQUmzdvxrJly7B161audj+t0JTHa4QkSTCbzVi/fj3WrFkDu92Ojo4OdHZ2Yu3ataipqUF5eflFT0jkV/fcfffdSCaTcDqdcDgcaGlpwf33349Vq1ahrKysYHdgYACNjY148MEHr+jm+/zFL34Rdrsddrsd77//Pr7yla9csc9msxn33HPPhXZdLhcaGhrw8MMPY/Xq1Vds98tf/jLuv/9+2O129PX14Z133sHGjRsv2+6nFQruNSZ/C9QFCxZgwoQJsFgsWLJkCZc7e/ZsTJgwAdFoFAsWLOB2p0yZAq/Xy+VarVYsXLgQc+fORTweF+pzTU0N+vv7hdpdsGABgsFgwe1+2qC3MILQIBRcgtAgFFyC0CAUXILQINzBHRgYKHv77bdv6erquvit7zmg0/oEIQb3WeWf/OQnt7/33nuznnjiCd0//uM/nlZVVcpfYOdFVVWIuvlbeIiWiLwR7Y7EvVHHaiTtarHPIx1fVVVH1OdC4Q6uoiisqqoqYDAYTJlMRrdr164lR48eLRXtbGNjI7cHfHiQjh07Jnxt79ixY9Dr9dxeOBzGwMAAfD4ftxuPx9Hd3S1cFKq9vV1ouVk2m8WpU6eEi4Ll75LBi6qqOHbsmNA3K8YYGhsbhcYoGAzC7XZjeHiY241Go7Db7QgGg9xuMpnE2bNnkUgkuN1MJsP1WuYO7vLly4PxeHx8ZWWl12QyqQ8//HCTqqqxjRs38u4KsixDr9fj0Ucf5XbzL6THHntMyFVVVch1u91obW3F6tWrud1AIICjR49i7dq13G4sFsOoUaNQX1/P7aZSKVgsFmzYsIHbzeVyMBqN2LRpE7ebX54oOkaMMSE3X+py5cqV3K7X60VzczPWrFnD7YbDYTQ0NGDdunXcbiKRwJtvvlnw9tzB/eY3v3nmm9/85hlejyCIqwedVSYIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1Cd3kkCA3CHdy9e/fOePrpp9e2tbUVfhfrS3Az3+WRIK4l3HOVt2/fXuN0Otmrr766aOHChe8MDAyM9vl8Rrvdzt24oijwer0QcRlj8Pl8Qi4AYdfn82F4eFjIDYfD8Hg8Qm4ikRA+VplMRtiVZVlzYzQ8PCw8RsFgUNiNxWLC45tKpZDL5Qrenju4Tz/99MmtW7feYzQafblcTnK5XJWBQMAwODjIuyuoqopgMAgRN38nfxEXgLAbDAbh8/mE3Gg0Kuwmk0n4/X4hN5vNIhAICLmKooxojETbBcTHyOfzCR/ncDgsfJzj8biwm06nuZZdcgd39OjR+rvvvtu1cuXKM0ajkS1durR7cHAw9bnPfY53V5BlGU6nEyIuYwz9/f3CrsPhEHLdbjeKi4uF3EAgAMaYkBuLxZBIJITcVCoFv98v5OZyObhcLiFXVVUMDAxc9zFyOp0oLy8Xcr1eLwwGg5AbDoeRzWaF3Gu+rK+mpsZTU1NTeOlygiCuOnQ5iCA0CAWXIDQIXcclCA1C9XEJQoPQJy5BaBD6xCUIDUInpwhCg1BwCUKD0G9cgtAgBQU3k8noGxoabvF4PJa33357TlNT07ir0Tj9xiUIMQqa8qgoCv793/995pNPPmnZsWPH2EwmU/Taa6+9ptPpmKqqOsYYFEXhbjxfIEnEBT4M/vV2R9Lnkboj6fONeL75Ui83Yoy0Or6FUlBwS0pKlBkzZrgBmJcsWeLv7++fCAAfFf26faRFv0SLQo2k6JdoQalIJIL+/n74/X5uN1/0KxqNcrsjKfqVy+XQ0tIyoqJfItyoMRpp0S+Hw4FQKMTtJpNJ2Gy2v56iX+l02mixWGZ0dHREHA6Hefz48S6dTsdMJhOjol+FcyOLfpWUlNyQol+SJF33MaKiXx9hNptz//mf//kid28Igrgm0FllgtAgNHOKIDQITcAgCA1CwSUIDULBJQgNQsElCA1CwSUIDULBJQgNQsElCA1CwSUIDVJQcGOxmPEHP/jB+tbW1gn/8A//8PCRI0fGXOuOEQRxaQqaq1xaWpobHBz0MsbM1dXV/nPnzlnvuusuHwD09/eP8Xq9xt7eXu7G80W/RFzGmLALfFhfRsT1+Xxwu91CbjgcxvDwsJCbSCTg8XiE3HzRLxE3X/RLxFVV9YaM0fDwsPAY5VcWibixWEx4fK9Z0S+TyWRMp9PmiooKYzabNQNALpeThoeHy0OhkGFoaIi7s6qqIhQKQcRljAm7AITdYDCIQCAg5EajUWE3mUwiGAwKudlsVtjNF/3S0hj5fD74/X4hNxwOC49RPB4XPlbXpOhXJpMxrFy5Uh0aGrJWVFSoo0aNsgCA0Whkd9xxR8/AwEBqxYoV3J2VZRmDg4MQcRlj6OvrE3btdruQ63a7YTabhdxAIABVVYXcWCyGeDwu5KZSKfh8PiE3l8thaGhIyFVVFf39/dd9jJxOJ8rKyoRcr9cLvV4v5IbDYWQyGSH3mizrM5lM8le+8pXD3L0hCOKaQGeVCUKDUHAJQoPQQnqC0CC0kJ4gNAh9VSYIDULBJQgNQsElCA1CwSUIDULBJQgNUnBwE4mEWZZlXTweNyeTSf66EARBXDUKCm4kEimqr6/fdOzYsZof/OAH/+fpp5++91p3jCCIS1PQXOXy8vLslClTbKNHj87W19fbfvWrX80DgGw2qzt06NDCU6dOWaxWK3fjqqri1KlTKCsr43YZYzh16hQqKiq4XQA4ffo0Kisrub1QKASHw4FMJsPtxmIx2Gw2oevXqVQKp0+fhtFo5Haz2SxaWlpQUlLC7SqKgpaWFpSWlnK7+TEqLy/ndgHxMQoEAnA6nUgmk9xuJBJBd3e3UIG0RCKB9vZ2oSJnmUyGq1hYQcHNZrP6eDw+5d133y393e9+t/Cpp556EwCKiorUJUuW9LpcrtTy5cu5O6soCjweD0RcxhhcLpeQC3y4gkTE9Xq9sFgsQm4oFIJOpxNy4/E40um0kJvJZBCJRIRcWZbh8/mEXFVV4Xa7hcdoaGhIyHW5XKisrBRy/X4/ioqKhNxoNApFUYTcZDKJDz74oODtCwquXq9Xf/rTn76nKApbtWpVp9lsvvDWUFFRkbBYLOqoUaO4OyvLMiwWC0RcxtgNcTOZDEpLS4XbtVqtQq7RaBR2U6mU8PPN5XLCrqqqIxqjkpISITeZTAofK1mWhcdXp9MJt2symbhKihYaXDZt2jQXd28Igrgm0OUggtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIBRcgtAgFFyC0CAUXILQIAUFN5vN6pqamub4/X5rU1PT7NbW1tFXo3G6yyNBiFHQlMdcLif98Ic/XPgv//Ivyrlz5yr27dt32+7du18BAMaY9NF/uRtnjF34E3FF2/3LfYh417tdLT7fkYzvn+9D1LkR4zvS51soBQXXYrEos2bNclZWVso1NTUBSZKmAUAmk9G9/vrrd3zwwQelIsugVFVFY2Oj8OAcPXqU2/tzV+QTPxwOY3BwEF6vl9uNx+Po7u5GKBTidtPpNNrb27mWfuXJZrM4deoUstkst6soCpqamqCqKrebH18RRjJG+Yp7Lhf/9PpoNAqHw4FAIMDtJpNJ2Gw2xGIxbjeTyXAtBywouOl02mgwGGr27NkzVqfTKTNnzgwAgMlkUuvr648zxmIbN27k7qwsyzAYDHj00Ue5XcYYJEnC448/LuQCEHLdbjdaW1uxevVqbjcQCODo0aNYu3YttxuLxXDw4EHU19dzu6lUClarFRs2bOB2c7kcioqKsGnTJm5XVVXo9Xo89thj3O5IxsjpdMJms2HlypXcrtfrRXNzM9asWcPthsNhNDQ0YN26ddzuNSn6ZTabc7/4xS9eyB9MvV5PdzIniBtIwfVxdTodhZUg/kqg2kEEoUGodhBBaBCagEEQGoSCSxAahH7jEoQGod+4BKFB6KsyQWgQCi5BaBAKLkFokIKCG4/Hjc8888xDHo9n1KFDh6b/+te/XnytO0YQxKUpaMqjyWSSOzs7Ix0dHWN/8YtfLFQUJfPkk0+2AIDb7a4IhUKS0+nkXnqiKAqCwaDkdDq5z1IxxvBRu0JnuERdr9cLv98v5IbDYQQCASE3kUgIH6tMJiPsyrI8ojESdQHxMXK73cJjFAgEhN1YLCY8vqlUSuVZYVdQcI1GI7NarTh8+PDEqqqqCV1dXfpgMKgrLS1lPT0944aGhl6y2WyneDurqqrO5XItstlsJ3ldxpjkcrk+a7PZmnldAHC5XLeLuMFgsGxwcHCCzWY7x+vGYjHL4ODgZ2w2Wwevm0qlzE6nc7bNZmvldbPZrHFoaKjWZrOd5nUVRdG7XK5bbTZbC687wjESdv1+f8XAwMBYm83WxeuGw2Hr4ODgFJvNdpbXTSQSxU6nc6bNZmvnddPpdFE2my0qWPjzxc6X+kulUobf/e53q3fv3n1HNBot/uMf/zg9/1g2m9W/+eabCwrZz1/+ybIs7d279xYRlzGG3bt333q9XZ/PZz1y5MhMETccDhc3NDTMFnHj8XjRoUOH5om46XTasH//fqExyuVyun379i0UPc579uy57uM7PDxc1tjYOF3EDQQCJe+///4sETcajZrefvvtuSJuMpk0vvXWW/ML3b6g37hms1n+2te+9tbatWubSktLUw8++KA9/1hDQ8Nkn89nSSQS3FXqm5qaPlNWViY0C2Pnzp0zwuFwSSgUKvxd6iN27do1ORQKWf1+v5nXLS4uTvT29pZFIhHuYrNer3eM2+2eGI1GLbyuz+fTOZ3OyiNHjozhdTs7Oyv8fv/4w4cPz+J133333c94vd6yvr4+7uP83nvvTfN6vcWRSKTgVWjpdNqwbdu2FQ6HozIUChXt2bNneqFuNpvVbd++/a5EImHq7OysPnPmTFWhrizL0iuvvHJnMBgsHxgYGLVz587Zhbqqqko7d+68w+12j4nFYsU7duyozWazBb+u//jHPy52Op3j582b17dt27ZFhTgjOqscDAb1mzdvvv3gwYNlx44dm8Drezye2Ouvv17wAfpzRo0a5XvnnXcqDxw4MJbXTSaTkS1btsxsaGgYz+u+8847k773ve993uv1cldrPnny5Iz333/fks1muW8n8atf/WrR3r17x/f19XG7s2bNCre1tcl2u52r+ngikZB++MMfLv23f/u3hX19fVwVppPJpO7HP/7xHb/97W/HHjx4sODwSZKkvv766yWtra21jY2NVS+++OLtiqIUFAJJktiePXuKBgYGxtnt9qI//elP4wpt12AwsEOHDum6u7snjBo1yv78888vTiaTBb3h6HQ69t5776nt7e2TDAZD9De/+U1dOp0u+M2qqakp19jYOP2///u/Fz3zzDN3+v3+Kz7fEQWXMSYVFRWxMWPGJBVF4S6VPnXq1JgkSYpI2wMDA5ZEImH6whe+wH0Pmdra2vTChQuHYrGYidd97733Pjtx4sSygYGBgl8UecaNG3fKaDRG//d//3cBrxuPx0snTJiQaW5u/iyvazAYlM7OzjGf+9znzvN4iqJIRqMxO3r0aHc2m+X6hlFSUqJ+61vfOm0ymapzuVzBL2KTyaROmDAhwhgzTJo0KVVdXV1wm0ajkU2aNCliNpvVO+64I8rTXwCYPHlyWK/Xs/b29plTpkzpKSkpKfhs0dSpU8M6nQ41NTWBj+pHF5ytqVOnRlpbW0va2tpusVqtE9xu9xW/CY4ouFVVVfKqVat6i4qKKhctWuTk9aPR6LSKiorKvr4+7k8vh8NRu2zZMikajXJ/3T18+HDN+PHji5cvX+7hdZ999tk93/nOd3bMmzevn9ft6uoaVV5eXnnffff18brf/va3WyorKw1Lly7t5nWDwaD+rrvu6ps6dWqSxysrK1P/5m/+ZnDx4sXq2LFjfbztjhs3Trd8+fL+FStW9BbqZDIZw6RJkyb19fWZ4vG4ddq0ab2F3nEll8vpx48fP7mlpWVab2/vZKvVOk1V1YI+rWVZ1lVVVU3t6OiY7nQ65y9btswoy3JBrqqqktVq/Uxvb+/k/fv3L6mrq3MZjcZcIS4AmEymz0yaNKl0+/btz3/961/fM3Xq1PSVHIkxmi9MEFqDZk4RhAah4BKEBqHgEoQGoeAShAah4BKEBqHgEoQGoeAShAah4BKEBqHgEoQG+X/Dcc4OaIWRHwAAAABJRU5ErkJggg=="
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -711,14 +734,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"Alice is trying to build a vertical line now. I need to block her next move otherwise she will have a chance to win. I will place my piece at [5, 7].\",\n",
       "    \"move\": [\n",
       "        5,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -739,8 +762,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -751,14 +776,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
+      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [8, 7] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        8,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
+      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -779,8 +804,10 @@
     },
     {
      "data": {
-      "text/plain": "<Figure size 432x288 with 1 Axes>",
-      "image/png": "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"
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
      },
      "metadata": {
       "needs_background": "light"
@@ -791,7 +818,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
+      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
       "    \"thought\": \"Alice is trying to build a vertical line again. I need to block her next move otherwise she will have a chance to win. I will place my piece at [9, 7].\",\n",
       "    \"move\": [\n",
       "        9,\n",
@@ -824,17 +851,13 @@
     "            msg = player(msg)\n",
     "            \n",
     "            i += 1"
-   ],
-   "metadata": {
-    "collapsed": false,
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:10:33.989202Z",
-     "start_time": "2024-03-27T08:09:59.446824Z"
-    }
-   }
+   ]
   },
   {
    "cell_type": "markdown",
+   "metadata": {
+    "collapsed": false
+   },
    "source": [
     "## Future Direction\n",
     "\n",
@@ -845,10 +868,7 @@
     "- Fine tune a model to improve the performance of the agent\n",
     "\n",
     "For complete code, we provide in [code/game_gomoku.py](./code/game_gomoku.py)."
-   ],
-   "metadata": {
-    "collapsed": false
-   }
+   ]
   }
  ],
  "metadata": {

From 6e9dca67edc51ae27d6239663103198a96f6bc7d Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 22 May 2024 20:45:57 +0800
Subject: [PATCH 29/55] revert unnecessary changes

---
 examples/game_gomoku/main.ipynb | 254 +++++++++++++++-----------------
 1 file changed, 117 insertions(+), 137 deletions(-)

diff --git a/examples/game_gomoku/main.ipynb b/examples/game_gomoku/main.ipynb
index 569321c15..04be0c07b 100644
--- a/examples/game_gomoku/main.ipynb
+++ b/examples/game_gomoku/main.ipynb
@@ -31,13 +31,6 @@
   {
    "cell_type": "code",
    "execution_count": 1,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:57.602994Z",
-     "start_time": "2024-03-27T08:09:57.565787Z"
-    },
-    "collapsed": false
-   },
    "outputs": [],
    "source": [
     "YOUR_MODEL_CONFIGURATION_NAME = \"{YOUR_MODEL_CONFIGURATION_NAME}\"\n",
@@ -46,31 +39,31 @@
     "    \n",
     "    # ...\n",
     "}"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:57.602994Z",
+     "start_time": "2024-03-27T08:09:57.565787Z"
+    }
+   }
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "## Step 1: Prepare a board for Gomoku\n",
     "\n",
     "First we create a `BoardAgent` class by inheriting from `AgentBase`, which manage the game board and update the game status as follows.\n",
     "\n",
     "To create a better visual experience, we also provide a function `board2img` to convert the board to an image. "
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   },
   {
    "cell_type": "code",
    "execution_count": 2,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:58.153932Z",
-     "start_time": "2024-03-27T08:09:57.571266Z"
-    },
-    "collapsed": false
-   },
    "outputs": [],
    "source": [
     "import numpy as np\n",
@@ -110,27 +103,26 @@
     "    plt.savefig(save_path, bbox_inches='tight', pad_inches=0.1)\n",
     "    plt.close(fig)  # Close the figure to free memory\n",
     "    return save_path"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:58.153932Z",
+     "start_time": "2024-03-27T08:09:57.571266Z"
+    }
+   }
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "The following code shows the implementation of the `BoardAgent` class. The agent manages the game board and updates the game status."
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.392069Z",
-     "start_time": "2024-03-27T08:09:58.159470Z"
-    },
-    "collapsed": false
-   },
    "outputs": [],
    "source": [
     "import numpy as np\n",
@@ -237,13 +229,18 @@
     "            else:\n",
     "                count = 0\n",
     "        return False\n"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.392069Z",
+     "start_time": "2024-03-27T08:09:58.159470Z"
+    }
+   },
+   "execution_count": 3
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "### Step2: Prepare a Gomoku Player Agent\n",
     "\n",
@@ -253,18 +250,13 @@
     "2. Within the agent, to enable the agent to think, we ask LLMs to respond in a dictionary format, which contains the thought and the move (\"thought\" field must come before \"move\"). To achieve this, we prepare a parsing function to extract the dictionary from response. \n",
     "\n",
     "The implementation of the Gomoku agent is as follows: "
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.397713Z",
-     "start_time": "2024-03-27T08:09:59.392729Z"
-    },
-    "collapsed": false
-   },
    "outputs": [],
    "source": [
     "import json\n",
@@ -335,7 +327,15 @@
     "        \n",
     "        # Hide thought from the response\n",
     "        return Msg(self.name, response[\"move\"], role=\"assistant\") \n"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.397713Z",
+     "start_time": "2024-03-27T08:09:59.392729Z"
+    }
+   },
+   "execution_count": 4
   },
   {
    "cell_type": "markdown",
@@ -346,25 +346,17 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:09:59.446367Z",
-     "start_time": "2024-03-27T08:09:59.398116Z"
-    },
-    "collapsed": false
-   },
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m2024-03-27 16:09:59.427\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models\u001b[0m:\u001b[36mread_model_configs\u001b[0m:\u001b[36m171\u001b[0m - \u001b[1mLoad configs for model wrapper: post_api\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.435\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m_create_monitor_table\u001b[0m:\u001b[36m341\u001b[0m - \u001b[1mInit [monitor_metrics] as the monitor table\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.435\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m_create_monitor_table\u001b[0m:\u001b[36m342\u001b[0m - \u001b[1mInit [monitor_metrics_quota_exceeded] as the monitor trigger\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.436\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.utils.monitor\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m311\u001b[0m - \u001b[1mSqliteMonitor initialization completed at [./runs/run_20240327-160958_kbf2ta/agentscope.db]\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.440\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models.model\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m257\u001b[0m - \u001b[1mInitialize model [post_api]\u001b[0m\n",
-      "\u001b[32m2024-03-27 16:09:59.441\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36magentscope.models.model\u001b[0m:\u001b[36m__init__\u001b[0m:\u001b[36m257\u001b[0m - \u001b[1mInitialize model [post_api]\u001b[0m\n"
+      "\u001B[32m2024-03-27 16:09:59.427\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models\u001B[0m:\u001B[36mread_model_configs\u001B[0m:\u001B[36m171\u001B[0m - \u001B[1mLoad configs for model wrapper: post_api\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.435\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m_create_monitor_table\u001B[0m:\u001B[36m341\u001B[0m - \u001B[1mInit [monitor_metrics] as the monitor table\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.435\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m_create_monitor_table\u001B[0m:\u001B[36m342\u001B[0m - \u001B[1mInit [monitor_metrics_quota_exceeded] as the monitor trigger\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.436\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.utils.monitor\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m311\u001B[0m - \u001B[1mSqliteMonitor initialization completed at [./runs/run_20240327-160958_kbf2ta/agentscope.db]\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.440\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models.model\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m257\u001B[0m - \u001B[1mInitialize model [post_api]\u001B[0m\n",
+      "\u001B[32m2024-03-27 16:09:59.441\u001B[0m | \u001B[1mINFO    \u001B[0m | \u001B[36magentscope.models.model\u001B[0m:\u001B[36m__init__\u001B[0m:\u001B[36m257\u001B[0m - \u001B[1mInitialize model [post_api]\u001B[0m\n"
      ]
     }
    ],
@@ -389,13 +381,18 @@
     ")\n",
     "\n",
     "board = BoardAgent(name=\"Host\")"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:09:59.446367Z",
+     "start_time": "2024-03-27T08:09:59.398116Z"
+    }
+   },
+   "execution_count": 5
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "### Step 3: Start the Gomoku game\n",
     "\n",
@@ -404,32 +401,26 @@
     "In this game, we use a message hub to share messages between the two players (board agent doesn't have memory), so that one player can hear what the board agent says to the other player, and what moves the other player makes. \n",
     "\n",
     "Note here we only show 10 steps of the game. You can adjust the `MAX_STEPS` to play more rounds."
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   },
   {
    "cell_type": "code",
    "execution_count": 6,
-   "metadata": {
-    "ExecuteTime": {
-     "end_time": "2024-03-27T08:10:33.989202Z",
-     "start_time": "2024-03-27T08:09:59.446824Z"
-    },
-    "collapsed": false
-   },
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: Welcome to the Gomoku game! Black player goes first. Please make your move.\n"
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: Welcome to the Gomoku game! Black player goes first. Please make your move.\n"
      ]
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -440,14 +431,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"As the first move of the game, it's best to place the piece in the center of the board to maximize possible winning directions.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -468,10 +459,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -482,14 +471,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"As the game has just started, I will place my piece near Alice's to prevent her from forming a line and to start building my own line.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        8\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -510,10 +499,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -524,14 +511,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"My opponent placed his piece next to mine. I should block him and also try to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        6\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -552,10 +539,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAO4AAADnCAYAAAAZ4WrqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAjBklEQVR4nO2de3BU5f3/33vLbrKbkHBP5CYQUAT8ChVBBH+2Iqgt/mhDK8VbW7W2OtQ6nbY67fQy+q3Wjq2tl2nHS1ssXsAqF0XwEkAghCRAEmBJQjbJZrOb7P1+9uyec57fH7j8bIu6z7NAPNPPaybjOOzrPM+ez773cs55zsfAGANBEPrCONwTIAiCHwouQegQCi5B6BAKLkHoEAouQegQCi5B6BAKLkHoEO7gHj9+vHL9+vWLBgcHy87FhAiC+GwMvBdgrFmzZllVVZVSU1OT/tnPftYoy7KZMWY6R/MjiP8azGazYjab1YIey7txRVHMF110kd/v94+QZdm4devWLxw+fPjh2bNnV/NuS9M0HD16FHPnzuVVwRhDW1sbLr30Um4XAFpbW4XcZDKJoaEhTJs2jduVJAn9/f2YMWMGtyvLMlwuFy6++GJuV1EUdHR04JJLLuF2NU3DsWPHMGfOHG6XMYb29vbzXt94PI5gMIipU6dyu6lUCl6vF7W1tdxuJpNBb28vLrroIm43l8ux0tLS/129evWGggTGGNffa6+9NuXOO+/8elNT0xjGGGRZtrz00kttTIBcLsf+/ve/i6hM0zT217/+Vdh98cUXhVyv18u2b98u5AaDQbZ582YhNx6Ps40bNwq56XSabdiwQcjNZrNs/fr1Qq6qquxvf/ubkFtMjfr7+9nOnTuF3KGhIbZt2zYhNxKJsH/+859CbjKZVF599dV7WIE55P7EXb16de/q1at7eT2CIM4edFSZIHTIsAbXYDAM5/AEoVuGNbiMlhQShBD0VZkgdAgFlyB0CAWXIHQIBZcgdAgFlyB0CAWXIHQIncclCB3CHdy9e/dO/NnPfrb2bAxO53EJQgzua5UnT57sczqdmqZpBk3TcPz48Qlut9t65MgR7sE1TYPb7YaIyxgTdgGgv79fyA0Gg3C5XEJuLBZDT0+PkJtOp9Hb2yvkZrNZ9PX1Cbmqqgq7w1WjoaEh4X0ViUSE65tMJoXrm8lkIMtywY/nDq7BYLDIslwRi8Vs5eXlGU3TDIwxaJrGuylomnb6TwTRcT8+tsiYouMW42qapju3mOeb9/+b9jMP3ME1Go22u+++uzMUCpVVVVVJs2fP7nc6nfK8efN4NwVFUeB0OiHiso/Weoq6bW1tQq7P54OqqkJuKBRCPB4XchOJBAYHB4VcSZLQ09Mj5OZyOXR0dAi5+bW857tGHo8HJpNJyPX7/chkMkJuNBpFKBQSclOpFDweT8GP5w5uTU1NZOXKlfW8HkEQZw86HUQQOoSCSxA6hIJLEDqEgksQOoSunCIIHUJ3wCAIHUJflQlCh1BwCUKHUHAJQodwBzcSiVj9fv8IRVHoyBJBDBPclzw+99xz1U6n8/KlS5ceueOOO7rOxaQIgvh0uIN722239d93332XVFRUpPNNvw4cOOAwmfgb9mmahgMHDsBisXC7jDE0NjbCarVyuwDQ2NgIm83G7UWjUfT19SEajXK7yWQSHR0dSKfT3K4kSWhra4OiKNxuNptFc3Oz0Ok3VVXR0NAAs5n7pXK6viUlJdxuvr4iNQqFQvB6vQiFQtxuPB5Hd3c3EokEt5tOp3Hs2DGu5Xl5ZFnmqi13NR555JEFpaWlU2bNmtVgtVq16667rjWZTKZuuukm3k1BVVVIkgQRlzGGZDIp5AKnVtuIuIODg2hvb8eyZcu43XA4jMrKStx4443cbiKRgNVqFZpzJpMBACFXURTIsizkapqGdDp93us7MDCAjo4OfPGLX+R2A4EAWlpasGLFCm43FovBbrdj5cqV3G46nca7775b8OO5g/urX/3qSDab7bDb7QkAsNlsOYvFwkpLS3k3BUVRUFJSAhGXMVaUa7FYhFybzSY8bjFuMfsKgPDzzeVywuNqmvZfVSNZlovaV0Zj4YecuINbVVUlAZB4PYIgzh50OoggdAgFlyB0CAWXIHQIBZcgdAgFlyB0CAWXIHQIBZcgdAgFlyB0CAWXIHQId3D37ds3etu2bfODwSD/1d8EQZwVuC95dLlc8vPPPz/j61//eub73//+MUVRTJqmCa1aUVUVqqoKuYwxYRc4dW3o+Z5zMa6iKEWNK/p8ixlX07Si6lvMnPVYX557sHEHd+HChVJTU5M/nU5bPlrWN7+hoaGcdzvA/1/2JbIkEDi1NE9kSSAA4eWEsVgMbrcb4XCY202lUujs7EQymeR2M5kM2tvbkc1mud1cLoeWlhbhZlQHDhzgugA+D2MMBw4cEFoSCIjXNxKJwOv1wu/3c7uJRAIulwuxWIzblSQJx48fhyTxX8rPW1fuPbp169YLTSZT1aJFi5xWq1VbtWrVQUVREmvX8rfMVRQFRqMRt956K7ebf0cWdVVVFXJ9Ph9aW1uFln2FQiHs27dPaNlXIpFAVVUV6urquF1JklBaWoo1a9Zwu7lcDiaTCbfccgu3m3+jON818ng8cDqdQksv/X4/mpqahJZeRqNR1NfXY9WqVdxuKpXCW2+9VfDjuYP7wAMPdAE4K3e+oPsqE4QYdF9lgtAhdDqIIHQIBZcgdAj1DiIIHUK/cQlCh9BXZYLQIRRcgtAhFFyC0CEUXILQIXRUmSB0CHdwY7GY8f777/+Kx+MZVezgdFSZIMTgvlZ58+bN019++eVJd955Z2t1dXW4u7t7nM/ns5w4cYJ7cE3T4PP5IOIyxoRd4FQPIBE3GAzC4/EIubFYDAMDA0JuKpWC1+sVcmVZFt5XqqrqrkZDQ0PCNQqHw8JuIpEQrq8kSVwrhLiDG41GJ0yfPn1MT09P9cyZM/uj0ag9lUqZRJa5aZqGZDIptEQOOPViFnVFx41Go0gkEkJuPB4XdiVJEp5zNpsVdlVVFXYZY0XVSNQtpkbFuKlUSnhfZTIZrmWX3MFdt27dB5dffvmxiy66KGmxWNj8+fNdLpcrc+WVV/JuCoqioKenByIuYwxdXV3Cbmdnp5Dr8/lgNpuF3FAohGw2K+QmEglEIhEhV5Ik+Hw+ITeXy6Gvr0/I1TQN3d3d571GHo8HZWVlQq7f7wdjTMiNRqNIJpNCbiqVQjAYLPjxQiucFy1aNCTiEQRxdqDTQQShQ+h0EEHoEFpkQBA6hL4qE4QOoeAShA6h4BKEDqHgEoQOoaPKBKFD6KgyQegQ7uD29fVVb9++fWEymSw9FxMiCOKz4b7kcc+ePZc0NTXZFi1axADgo6Zfhlwuxz14vkmSiJtvUSHi5scWcfNNsMj9bPJNv0RrpGmarp5vse45bfo1YcKEw3v37p33/PPPz7rvvvuOfNT0yyHyezXfUEqkKRRjDI2NjSgpKeF2AeDgwYOwWq3cXjQahdvtRiQS4XaTySQ6OzuRSqW43Uwmg7a2NqEXRS6XQ3NzM7cHnHqDE23Mlq+RaGM20fqGw2F4vV6ui/bzxONxuFwuxONxbjedTuP48ePIZDLcrizLXI/nTsyRI0cq7Xb7iEWLFh07G02/DAaDUEOpfNMvUVdRFCF3OJt+VVZWCjf9stlsw9L0izF23mtETb/OwA9/+MNuAN28HkEQZw86j0sQOoSCSxA6hIJLEDqEgksQOoSCSxA6hIJLEDqEgksQOoSCSxA6hIJLEDqEgksQOoQ7uLt27Rrzk5/8ZOUbb7wx9lxMiCCIz4b7WuU//vGPCw0GQ+lll11mUFXVcPLkyfFer7fE6XRyD66qKrxeL0RcxpiwC5xaLCDiBgIBuN1uITcajaK/v1/ITaVSpy+e50WWZQwMDAi5iqII72dN04qqr2iNBgcHhWsUDoeFa5RIJIRrdM6bfk2aNEkGkDl69OgldXV1/kQiUSpJkjEWi/FuCpqmIZ1OQ8RljEGSJCEXgLCbSCSE51yMm06nheeczWaFXVVVqUYFkkwmhd3z0fSrobm5+ZLa2lqn2Wxml112Wc/JkyczCxcu5N0UFEVBd3c3RNx8UyhR98SJE0Kuz+eDyWQSckOhEDKZjJCb7yAn4kqShIGBASE3l8uhp6dHyNU0DV1dXee9Rh6PB6WlpUKu3++HqqpCbr7Tn4ibSqXg9/sLfjx3cKdOnZqYOnXqAV6PIIizBx1VJggdQsElCB1CwSUIHULBJQgdItSR/mxBnQw+HcYYcrkc/H4/+vr60NfXB5fLherqathstk/dfx93PR4P3G43t+v1euHxeNDV1YWOjg5MnDgRpaWlVLfPAcMaXOpkcGby94xubm5GZ2cnxo4di3HjxuFLX/oSent7sX//fowZMwZLliz5jyDl747Y0tLyL+7y5ctPu2PHjsVVV111xhDmcjns2rULzzzzDA4ePIhAIABVVfHMM89g9uzZ+M53voOvfvWrFOBhZliDS5yZTCaDbdu2oaamBt/4xjdQUlJyOiT5UHd2dmLTpk244YYbMGrUqNP/Xojb0dFxRleSJDz22GP4/e9//x/3FQ4Gg9i1axcaGhrwzjvv4IknnsCYMWMovMME/cb9nKGqKnbu3IlZs2bhyiuvhNVq/ZdwGAwGmM1mXHzxxbjuuuuwffv20zfTVlUVO3bs+Ex31qxZZ3SfffZZPProo596M3BZlvHyyy/jpz/9KfdNvImzBwX3cwRjDD09PafD9WmfZgaDAePGjcOcOXNw4MABMMbgcrlQUlLC5TY2NoIxho6ODjz++OMFhVHTNLzyyit455136OfOMMEd3EQiYYlEIuWSJIn1lSA+ldbWVixYsKCgr6AGgwGzZs2C2+2GoijCbi6Xw6ZNmzA0NFTwPCVJwvr164X7AhHFwR3ctra2cffcc883/vGPf8w+FxP6byZ/Qf+oUaMKdiwWCyoqKhAMBpHJZDBy5Egu1+FwIBQKobm5mfvT8/Dhw0gkElwOcXbgPji1ePFiz29/+9vokiVLumVZNm7fvv2ygwcPOkQaaGmahqamJpSW8nfsZIyhqakJdrud2wWA5uZmOBwObi8ajaK3txfJZJLbTSaTOHHixCcu35JlGfF4HEZj4e+nBoMBFosFGzduhMVi4TpYZDAYYDKZsHHjRuFlips2bfrEN5p8jcrKyri3DYjXKBQKYWBgQGiVTjwex8mTJyFJErebTqdx9OhRqKrK7cqyzHXMgDu4nZ2dDpvNpsycOTMOAEuXLnWGQqH08uXLeTcFVVURi8Ug4jLGEA6HhVzg1FFSEXdoaAjt7e249tprud1wOIzS0tJPHFdRFGzevBmaphUc3vw51xUrVuDAAb61H/mjzNdffz3efvttnDx5kssvLy/HihUrPvFTXtM0RCIR4fqK1mhgYACdnZ245ppruN1gMIiKigqhcWOxGMxms5CbTqdRX19f8OO5g3vBBRek//SnP72T/3+Hw5Gx2WxaeXk576agKApsNhtEXMbYsLjJZBKlpaVCbjab/VSXMQa73Y5oNFrwV15FURCPxzFlyhQcOXIEsVgMlZWVBbvJZBKTJk3C/PnzsWPHjkKfCgBgzpw5GDduHGw22xn/XdO0YamRw+EQrpEkScKuqqrCrtFo5Pqmxf0b1263a2PHjuVvAHoG6BzgfzJ79my0tLQU/Hvz5MmTqK6uhsViwezZs9HU1FSQyxhDV1cXqqurUVJSgrq6Oq7f1iUlJVi7dq1Qj2GieIb1dBCdSvhXDAYDamtrEY1G4XK5PnX/5H8qHDx4EFdeeSUMBgNmzJhRsBuJRNDU1IRFixbBYDBgzpw5uPfeewtqQm0wGLBy5Up85StfoTffYWJYg0tF/0/MZjOuv/56HDhwAEeOHIGiKP8SwnxD776+PmzZsgXLli07ffDn311VVc/o9vb2ntH90Y9+hO9973uf+NX342M88cQTwgcGieKha5U/h9jtdtTV1WHfvn145ZVXMGnSJFRXV8NkMiEQCKC3txdWqxWrVq1CRUXFv7wBftx9+eWXuVyHw4HHHnsMixcvxtNPP4329nbEYjEwxuBwODBt2jTccccduP322zFixAh64x1G6FrlzyEGgwFWqxXXXHMN0uk0PB4Penp60NLSgi9/+ctYvnw5Kioqzngw40yu2+1GQ0MDbrrpps90bTYbVq9ejRtuuAEulwsulwt79uzBzTffjNraWowYMYLrIApxbqDgfo4xGAyw2+2YOXMmampqEI/HMWfOHG530qRJ8Pv9XK7D4cDcuXNx8cUXI5lMYsGCBcU8FeIsQ2+dBKFDKLgEoUMouAShQyi4BKFDuIPrdrsr3nvvvUs7OzvFrhz/GHQ6gSDE4D6q/Itf/OLy3bt3z7jnnnuMP/7xjw9rmmbIn9jnRdM0iLqMMWE375/vcYtxh2tfFTOuHudcbH01TStqzoXCHVxVVdmoUaNCZrPZKsuy8Y033liwb9++ctHJNjQ0cHvAqZ20f/9+4XOK+/fvh8lk4vai0SjcbjcCgQC3m0wm0dXVJdwUqr29XWi5WTabxaFDh6AoCrerqurpu2Twomka9u/fL/TNijGGhoYGoRqFw2H4fD4MDg5yu/F4HC6XC+FwmNtNp9M4fvw4UqkUtyvLMtdrmTu4S5YsCSeTyfFVVVV+q9Wqfe1rX2vUNC2xdu1a3k1BURSYTCbceuut3G7+hXTbbbcJuZqmCbk+nw+tra1YsWIFtxsKhbBv3z6sXLmS200kEhg5ciTq6uq4XUmSYLfbsWbNGm43l8vBYrHglltu4XbzyxNFa8QYE3LzrS6XLVvG7fr9fjQ1NeHGG2/kdqPRKOrr67Fq1SpuN5VK4a233ir48dzBveuuu47cddddR3g9giDOHnRUmSB0CAWXIHQIBZcgdAgFlyB0CAWXIHQIBZcgdAgFlyB0CAWXIHQIBZcgdAjd5ZEgdAh3cLdu3Tpt3bp1K9va2qqKHZzu8kgQYnBfq7xhw4Zaj8fDXnvttXlz58593+12jw4EAhaXy8U9uKqq8Pv9EHEZYwgEAkIuAGE3EAhgcHBQyI1GoxgaGhJyU6mU8L6SZVnYVRRFdzUaHBwUrlE4HBZ2E4mEcH0lSeJqWcod3HXr1jU/++yzX7RYLIFcLmfwer1VoVDI3N/fz7spaJqGcDgMETd/J38RF4CwGw6HEQgEhNx4PC7sptNpBINBITebzSIUCgm5qqoWVSPRcQHxGgUCAeH9HI1GhfdzMpkUdjOZDNeyS+7gjh492nTNNdd4ly1bdsRisbCFCxd29ff3S1dffTXvpqAoCjweD0Rcxhj6+vqE3Z6eHiHX5/OhtLRUyA2FQmCMCbmJRAKpVErIlSQJwWBQyM3lcvB6vUKupmlwu93nvUYejwcjRowQcv1+P8xms5AbjUaRzWaF3HO+rK+2tnaotra28NblBEGcdeh0EEHoEAouQegQOo9LEDqE+uMShA6hT1yC0CH0iUsQOoQOThGEDqHgEoQOod+4BKFDCgquLMum+vr6S4eGhuzvvffeRY2NjePOxuD0G5cgxCjokkdVVfGb3/xm+r333mt/9dVXx8qyXLJx48aNRqORaZpmZIxBVVXuwfMNkkRc4FTwz7dbzJyLdYuZ83A833yrl+GokV7rWygFBbesrEydNm2aD4BtwYIFwb6+vgsA4KOmX5cX2/RLtClUMU2/RBtKxWIx9PX1IRgMcrv5pl/xeJzbLabpVy6XQ0tLS1FNv0QYrhoV2/Srp6cHkUiE202n03A6nZ+fpl+ZTMZit9unHT16NNbT02MbP36812g0MqvVyqjpV+EMZ9OvsrKyYWn6ZTAYznuNqOnXR9hsttzvfve79dyzIQjinEBHlQlCh9CVUwShQ+gCDILQIRRcgtAhFFyC0CEUXILQIRRcgtAhFFyC0CEUXILQIRRcgtAhBQU3kUhYHnzwwdWtra01P/jBD762d+/eMed6YgRBfDIFXatcXl6e6+/v9zPGbNXV1cETJ044rrrqqgAA9PX1jfH7/Zbu7m7uwfNNv0RcxpiwC5zqLyPiBgIB+Hw+ITcajWJwcFDITaVSGBoaEnLzTb9E3HzTLxFX07RhqdHg4KBwjfIri0TcRCIhXN9z1vTLarVaMpmMrbKy0pLNZm0AkMvlDIODgyMikYh5YGCAe7KapiESiUDEZYwJuwCE3XA4jFAoJOTG43FhN51OIxwOC7nZbFbYzTf90lONAoEAgsGgkBuNRoVrlEwmhffVOWn6JcuyedmyZdrAwICjsrJSGzlypB0ALBYLu+KKK0663W5p6dKl3JNVFAX9/f0QcRlj6O3tFXZdLpeQ6/P5YLPZhNxQKARN04TcRCKBZDIp5EqShEAgIOTmcjkMDAwIuZqmoa+v77zXyOPxoKKiQsj1+/0wmUxCbjQahSzLQu45WdZntVqVm2+++QPu2RAEcU6go8oEoUMouAShQ2ghPUHoEFpITxA6hL4qE4QOoeAShA6h4BKEDqHgEoQOoeAShA4pOLipVMqmKIoxmUza0uk0f18IgiDOGgUFNxaLldTV1d2yf//+2gcffPD/rFu37rpzPTGCID6Zgq5VHjFiRHbSpEnO0aNHZ+vq6pxPPfXULADIZrPGnTt3zj106JDd4XBwD65pGg4dOoSKigpulzGGQ4cOobKyktsFgMOHD6Oqqorbi0Qi6OnpgSzL3G4ikYDT6RQ6fy1JEg4fPgyLxcLtZrNZtLS0oKysjNtVVRUtLS0oLy/ndvM1GjFiBLcLiNcoFArB4/EgnU5zu7FYDF1dXUIN0lKpFNrb24WanMmyzNUsrKDgZrNZUzKZnLRr167yF154Ye599933FgCUlJRoCxYs6PZ6vdKSJUu4J6uqKoaGhiDiMsbg9XqFXODUChIR1+/3w263C7mRSARGo1HITSaTyGQyQq4sy4jFYkKuoigIBAJCrqZp8Pl8wjUaGBgQcr1eL6qqqoTcYDCIkpISITcej0NVVSE3nU7jww8/LPjxBQXXZDJpjzzyyG5VVdny5cuP2Wy2028NlZWVKbvdro0cOZJ7soqiwG63Q8RljA2LK8syysvLhcd1OBxCrsViEXYlSRJ+vrlcTtjVNK2oGpWVlQm56XRaeF8piiJcX6PRKDyu1WrlailaaHDZlClTvNyzIQjinECngwhCh1BwCUKHUHAJQodQcAlCh1BwCUKHUHAJQodQcAlCh1BwCUKHUHAJQocUFNxsNmtsbGy8KBgMOhobG2e2traOPhuD010eCUKMgi55zOVyhoceemjuL3/5S/XEiROV27Ztu2zz5s2vAABjzPDRf7kHZ4yd/hNxRcf9922IeOd7XD0+32Lq+/FtiDrDUd9in2+hFBRcu92uzpgxw1NVVaXU1taGDAbDFACQZdn4+uuvX/Hhhx+WiyyD0jQNDQ0NwsXZt28ft/dxV+QTPxqNor+/H36/n9tNJpPo6upCJBLhdjOZDNrb27mWfuXJZrM4dOgQstkst6uqKhobG6FpGrebr68IxdQo33HP6+W/vD4ej6OnpwehUIjbTafTcDqdSCQS3K4sy1zLAQsKbiaTsZjN5totW7aMNRqN6vTp00MAYLVatbq6ugOMscTatWu5J6soCsxmM2699VZulzEGg8GA22+/XcgFIOT6fD60trZixYoV3G4oFMK+ffuwcuVKbjeRSGDHjh2oq6vjdiVJgsPhwJo1a7jdXC6HkpIS3HLLLdyupmkwmUy47bbbuN1iauTxeOB0OrFs2TJu1+/3o6mpCTfeeCO3G41GUV9fj1WrVnG756Tpl81myz355JN/z+9Mk8lEdzIniGGk4P64RqORwkoQnxOodxBB6BDqHUQQOoQuwCAIHULBJQgdQr9xCUKH0G9cgtAh9FWZIHQIBZcgdAgFlyB0SEHBTSaTlocffvj/Dg0Njdy5c+fUp59+ev65nhhBEJ9MQZc8Wq1W5dixY7GjR4+OffLJJ+eqqirfe++9LQDg8/kqI5GIwePxcC89UVUV4XDY4PF4uI9SMcbw0bhCR7hEXb/fj2AwKORGo1GEQiEhN5VKCe8rWZaFXUVRiqqRqAuI18jn8wnXKBQKCbuJREK4vpIkaTwr7AoKrsViYQ6HAx988MEFo0aNquns7DSFw2FjeXk5O3ny5LiBgYF/OJ3OQ7yT1TTN6PV65zmdzmZelzFm8Hq9X3A6nU28LgB4vd7LRdxwOFzR399f43Q6T/C6iUTC3t/ff6HT6TzK60qSZPN4PDOdTmcrr5vNZi0DAwOznU7nYV5XVVWT1+v9H6fT2cLrFlkjYTcYDFa63e6xTqezk9eNRqOO/v7+SU6n8zivm0qlSj0ez3Sn09nO62YymZJsNltSsPDxxc6f9CdJkvmFF15YsXnz5ivi8Xjpm2++OTX/b9ls1vTWW2/NKWQ7//6nKIph69atl4q4jDFs3rz5f863GwgEHHv37p0u4kaj0dL6+vqZIm4ymSzZuXPnLBE3k8mY3377baEa5XI547Zt2+aK7uctW7ac9/oODg5WNDQ0TBVxQ6FQ2Z49e2aIuPF43Pree+9dLOKm02nLO++8c0mhjy/oN67NZlO+9a1vvbNy5crG8vJy6aabbnLl/62+vn5iIBCwp1Ip7i71jY2NF1ZUVAhdhbFp06Zp0Wi0LBKJFP4u9RFvvPHGxEgk4ggGgzZet7S0NNXd3V0Ri8W4m836/f4xPp/vgng8bud1A4GA0ePxVO3du3cMr3vs2LHKYDA4/oMPPpjB6+7atetCv99f0dvby72fd+/ePcXv95fGYrGCV6FlMhnzSy+9tLSnp6cqEomUbNmyZWqhbjabNW7YsOGqVCplPXbsWPWRI0dGFeoqimJ45ZVXrgyHwyPcbvfITZs2zSzU1TTNsGnTpit8Pt+YRCJR+uqrr87OZrMFv67ffPPN+R6PZ/ysWbN6X3rppXmFOEUdVQ6Hw6bHH3/88h07dlTs37+/htcfGhpKvP766wXvoI8zcuTIwPvvv1+1ffv2sbxuOp2OPfPMM9Pr6+vH87rvv//+hAceeOBLfr+fu1tzc3PztD179tiz2Sz37SSeeuqpeVu3bh3f29vL7c6YMSPa1tamuFwuru7jqVTK8NBDDy389a9/Pbe3t5erw3Q6nTb+/Oc/v+K5554bu2PHjoLDZzAYtNdff72stbV1dkNDw6j169dfrqpqQSEwGAxsy5YtJW63e5zL5Sp59913xxU6rtlsZjt37jR2dXXVjBw50vXiiy/OT6fTBb3hGI1Gtnv3bq29vX2C2WyO//nPf16UyWQKfrNqbGzMNTQ0TP3DH/4w7+GHH74yGAx+5vMtKriMMUNJSQkbM2ZMWlVV7lbpkydPThgMBlVkbLfbbU+lUtZrr72W+x4ys2fPzsydO3cgkUhYed3du3d/4YILLqhwu90FvyjyjBs37pDFYon/5S9/mcPrJpPJ8pqaGrmpqekLvK7ZbFaPHTs25uqrr+7g8VRVNVgsluzo0aN92WyW6xtGWVmZdvfddx+2Wq3VuVyu4Bex1WrVampqYowx84QJE6Tq6uqCx7RYLGzChAkxm82mXXHFFXGe+QLAxIkToyaTibW3t0+fNGnSybKysoKPFk2ePDlqNBpRW1sb+qh/dMHZmjx5cqy1tbWsra3tUofDUePz+T7zm2BRwR01apSyfPny7pKSkqp58+Z5eP14PD6lsrKyqre3l/vTq6enZ/bixYsN8Xic++vuBx98UDt+/PjSJUuWDPG6jz766Jb777//1VmzZvXxup2dnSNHjBhRdf311/fyut/97ndbqqqqzAsXLuzidcPhsOmqq67qnTx5cprHq6io0L75zW/2z58/Xxs7dmyAd9xx48YZlyxZ0rd06dLuQh1Zls0TJkyY0Nvba00mk44pU6Z0F3rHlVwuZxo/fvzElpaWKd3d3RMdDscUTdMK+rRWFMU4atSoyUePHp3q8XguWbx4sUVRlIJcTdMMDofjwu7u7olvv/32gkWLFnktFkuuEBcArFbrhRMmTCjfsGHDi9/+9re3TJ48OfNZjoExul6YIPQGXTlFEDqEgksQOoSCSxA6hIJLEDqEgksQOoSCSxA6hIJLEDqEgksQOoSCSxA65P8BmEt6Ngtc35AAAAAASUVORK5CYII="
      },
      "metadata": {
       "needs_background": "light"
@@ -566,14 +551,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"I see Alice is trying to create a line horizontally. I need to block her next move otherwise she will have a chance to win. I will place my piece at [7, 5].\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        5\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -594,10 +579,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -608,14 +591,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [7, 9] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        9\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -636,10 +619,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -650,14 +631,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"Alice is trying to create a line horizontally again. I need to block her next move otherwise she will have a chance to win. I will place my piece at [7, 10].\",\n",
       "    \"move\": [\n",
       "        7,\n",
       "        10\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -678,10 +659,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -692,14 +671,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [6, 7] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        6,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -720,10 +699,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -734,14 +711,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"Alice is trying to build a vertical line now. I need to block her next move otherwise she will have a chance to win. I will place my piece at [5, 7].\",\n",
       "    \"move\": [\n",
       "        5,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -762,10 +739,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -776,14 +751,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[32m\u001b[1mAlice\u001b[0m\u001b[32m\u001b[0m: {\n",
+      "\u001B[32m\u001B[1mAlice\u001B[0m\u001B[32m\u001B[0m: {\n",
       "    \"thought\": \"My opponent is trying to block me. I should place my piece at [8, 7] to create a potential line for myself.\",\n",
       "    \"move\": [\n",
       "        8,\n",
       "        7\n",
       "    ]\n",
       "}\n",
-      "\u001b[36m\u001b[1mHost\u001b[0m\u001b[36m\u001b[0m: The current board is as follows:\n",
+      "\u001B[36m\u001B[1mHost\u001B[0m\u001B[36m\u001B[0m: The current board is as follows:\n",
       "000000000000000\n",
       "000000000000000\n",
       "000000000000000\n",
@@ -804,10 +779,8 @@
     },
     {
      "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
+      "text/plain": "<Figure size 432x288 with 1 Axes>",
+      "image/png": "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"
      },
      "metadata": {
       "needs_background": "light"
@@ -818,7 +791,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[35m\u001b[1mBob\u001b[0m\u001b[35m\u001b[0m: {\n",
+      "\u001B[35m\u001B[1mBob\u001B[0m\u001B[35m\u001B[0m: {\n",
       "    \"thought\": \"Alice is trying to build a vertical line again. I need to block her next move otherwise she will have a chance to win. I will place my piece at [9, 7].\",\n",
       "    \"move\": [\n",
       "        9,\n",
@@ -851,13 +824,17 @@
     "            msg = player(msg)\n",
     "            \n",
     "            i += 1"
-   ]
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2024-03-27T08:10:33.989202Z",
+     "start_time": "2024-03-27T08:09:59.446824Z"
+    }
+   }
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "collapsed": false
-   },
    "source": [
     "## Future Direction\n",
     "\n",
@@ -868,7 +845,10 @@
     "- Fine tune a model to improve the performance of the agent\n",
     "\n",
     "For complete code, we provide in [code/game_gomoku.py](./code/game_gomoku.py)."
-   ]
+   ],
+   "metadata": {
+    "collapsed": false
+   }
   }
  ],
  "metadata": {

From 46233ffb87a1f55c6545e16577be39f98a9feb18 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Fri, 24 May 2024 12:12:58 +0800
Subject: [PATCH 30/55] optimized data filtering speed (data preprocessing)

---
 .../huggingface_model.py                      |  44 +-
 .../distributed_simulation/run_simlation.sh   |  25 +
 src/agentscope/server/__init__.py             |  10 +
 src/agentscope/server/launcher.py             | 449 ++++++++++++++++++
 src/agentscope/server/servicer.py             | 313 ++++++++++++
 5 files changed, 801 insertions(+), 40 deletions(-)
 create mode 100644 examples/distributed_simulation/run_simlation.sh
 create mode 100644 src/agentscope/server/__init__.py
 create mode 100644 src/agentscope/server/launcher.py
 create mode 100644 src/agentscope/server/servicer.py

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index c5c2cbf3e..809c1fd1a 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -385,42 +385,6 @@ def fine_tune(
             )
             raise
 
-    def filer_sequence_lengths(
-        self,
-        max_input_seq_length: int,
-        dataset_obj: List[Dict[str, List[str]]],
-    ) -> List[int]:
-        """
-        Identifies and returns the indices of conversation
-          entries that exceed max_input_seq_length characters in length.
-
-        Args:
-            dataset_obj (List[Dict[str, List[str]]]): A list where
-                each dictionary contains 'conversations',
-                a list of two strings (question and answer).
-
-        Returns:
-            List[int]: Indices of conversations where the combined
-            length of the question and answer exceeds
-            max_input_seq_length characters.
-        """
-        # Initialize a list to store the sequence lengths
-        sequence_lengths = []
-
-        # list of indices that are too long
-        too_long = []
-
-        # Loop over the dataset and get the lengths of text sequences
-        for idx, example in enumerate(dataset_obj):
-            sequence_length = len(
-                example["conversations"][0] + example["conversations"][1],
-            )
-            sequence_lengths.append(sequence_length)
-            if sequence_length > max_input_seq_length:
-                too_long.append(idx)
-
-        return too_long
-
     def formatting_prompts_func(
         self,
         example: Dict[str, List[List[str]]],
@@ -495,10 +459,10 @@ def fine_tune_training(
 
         dataset = load_dataset(data_path, split="train", token=token)
 
-        indexes_to_drop = self.filer_sequence_lengths(300, dataset)
-
-        dataset_reduced = dataset.select(
-            i for i in range(len(dataset)) if i not in set(indexes_to_drop)
+        # filter out input sequences that are longer than certain threshold
+        dataset_reduced = dataset.filter(
+            lambda x: len(x["conversations"][0] + x["conversations"][1])
+            <= 1000,
         )
 
         formatted_dataset = dataset_reduced.train_test_split(test_size=0.1)
diff --git a/examples/distributed_simulation/run_simlation.sh b/examples/distributed_simulation/run_simlation.sh
new file mode 100644
index 000000000..6fac7c4c4
--- /dev/null
+++ b/examples/distributed_simulation/run_simlation.sh
@@ -0,0 +1,25 @@
+#!/bin/bash
+
+# default values
+base_port=12330
+hosts="localhost" # or "server1 server2 server3 ..."
+moderator_per_host=4
+model_per_host=8
+agent_type="random" # or "llm"
+max_value=100
+
+# check server-per-host
+if ! [[ "$1" =~ ^[0-9]+$ ]]; then
+    echo "Usage: $0 <server-per-host> <participant-num>"
+    exit 1
+fi
+
+# check participant-num
+if ! [[ "$2" =~ ^[0-9]+$ ]]; then
+    echo "Usage: $0 <server-per-host> <participant-num>"
+    exit 1
+fi
+
+mkdir -p log
+
+python main.py --role main --hosts ${hosts} --base-port ${base_port} --participant-num $2 --server-per-host $1 --model-per-host ${model_per_host} --moderator-per-host ${moderator_per_host} --agent-type ${agent_type} --max-value ${max_value}
diff --git a/src/agentscope/server/__init__.py b/src/agentscope/server/__init__.py
new file mode 100644
index 000000000..8b69a542a
--- /dev/null
+++ b/src/agentscope/server/__init__.py
@@ -0,0 +1,10 @@
+# -*- coding: utf-8 -*-
+"""Import all server related modules in the package."""
+from .launcher import RpcAgentServerLauncher, as_server
+from .servicer import AgentServerServicer
+
+__all__ = [
+    "RpcAgentServerLauncher",
+    "AgentServerServicer",
+    "as_server",
+]
diff --git a/src/agentscope/server/launcher.py b/src/agentscope/server/launcher.py
new file mode 100644
index 000000000..ed5ed7f67
--- /dev/null
+++ b/src/agentscope/server/launcher.py
@@ -0,0 +1,449 @@
+# -*- coding: utf-8 -*-
+""" Server of distributed agent"""
+import os
+from multiprocessing import Process, Event, Pipe
+from multiprocessing.synchronize import Event as EventClass
+import asyncio
+import signal
+import argparse
+from typing import Type
+from concurrent import futures
+from loguru import logger
+
+try:
+    import grpc
+    from agentscope.rpc.rpc_agent_pb2_grpc import (
+        add_RpcAgentServicer_to_server,
+    )
+except ImportError as import_error:
+    from agentscope.utils.tools import ImportErrorReporter
+
+    grpc = ImportErrorReporter(import_error, "distribute")
+    add_RpcAgentServicer_to_server = ImportErrorReporter(
+        import_error,
+        "distribute",
+    )
+
+import agentscope
+from agentscope.server.servicer import AgentServerServicer
+from agentscope.agents.agent import AgentBase
+from agentscope.utils.tools import (
+    _get_timestamp,
+    check_port,
+)
+
+
+def _setup_agent_server(
+    host: str,
+    port: int,
+    server_id: str,
+    init_settings: dict = None,
+    start_event: EventClass = None,
+    stop_event: EventClass = None,
+    pipe: int = None,
+    local_mode: bool = True,
+    max_pool_size: int = 8192,
+    max_timeout_seconds: int = 1800,
+    custom_agents: list = None,
+) -> None:
+    """Setup agent server.
+
+    Args:
+        host (`str`, defaults to `"localhost"`):
+            Hostname of the agent server.
+        port (`int`):
+            The socket port monitored by the agent server.
+        server_id (`str`):
+            The id of the server.
+        init_settings (`dict`, defaults to `None`):
+            Init settings for agentscope.init.
+        start_event (`EventClass`, defaults to `None`):
+            An Event instance used to determine whether the child process
+            has been started.
+        stop_event (`EventClass`, defaults to `None`):
+            The stop Event instance used to determine whether the child
+            process has been stopped.
+        pipe (`int`, defaults to `None`):
+            A pipe instance used to pass the actual port of the server.
+        local_mode (`bool`, defaults to `None`):
+            Only listen to local requests.
+        max_pool_size (`int`, defaults to `8192`):
+            Max number of agent replies that the server can accommodate.
+        max_timeout_seconds (`int`, defaults to `1800`):
+            Timeout for agent replies.
+        custom_agents (`list`, defaults to `None`):
+            A list of custom agent classes that are not in `agentscope.agents`.
+    """
+    asyncio.run(
+        _setup_agent_server_async(
+            host=host,
+            port=port,
+            server_id=server_id,
+            init_settings=init_settings,
+            start_event=start_event,
+            stop_event=stop_event,
+            pipe=pipe,
+            local_mode=local_mode,
+            max_pool_size=max_pool_size,
+            max_timeout_seconds=max_timeout_seconds,
+            custom_agents=custom_agents,
+        ),
+    )
+
+
+async def _setup_agent_server_async(
+    host: str,
+    port: int,
+    server_id: str,
+    init_settings: dict = None,
+    start_event: EventClass = None,
+    stop_event: EventClass = None,
+    pipe: int = None,
+    local_mode: bool = True,
+    max_pool_size: int = 8192,
+    max_timeout_seconds: int = 1800,
+    custom_agents: list = None,
+) -> None:
+    """Setup agent server in an async way.
+
+    Args:
+        host (`str`, defaults to `"localhost"`):
+            Hostname of the agent server.
+        port (`int`):
+            The socket port monitored by the agent server.
+        server_id (`str`):
+            The id of the server.
+        init_settings (`dict`, defaults to `None`):
+            Init settings for agentscope.init.
+        start_event (`EventClass`, defaults to `None`):
+            An Event instance used to determine whether the child process
+            has been started.
+        stop_event (`EventClass`, defaults to `None`):
+            The stop Event instance used to determine whether the child
+            process has been stopped.
+        pipe (`int`, defaults to `None`):
+            A pipe instance used to pass the actual port of the server.
+        local_mode (`bool`, defaults to `None`):
+            If `True`, only listen to requests from "localhost", otherwise,
+            listen to requests from all hosts.
+        max_pool_size (`int`, defaults to `8192`):
+            The max number of agent reply messages that the server can
+            accommodate. Note that the oldest message will be deleted
+            after exceeding the pool size.
+        max_timeout_seconds (`int`, defaults to `1800`):
+            Maximum time for reply messages to be cached in the server.
+            Note that expired messages will be deleted.
+        custom_agents (`list`, defaults to `None`):
+            A list of custom agent classes that are not in `agentscope.agents`.
+    """
+    from agentscope._init import init_process
+
+    if init_settings is not None:
+        init_process(**init_settings)
+    servicer = AgentServerServicer(
+        host=host,
+        port=port,
+        max_pool_size=max_pool_size,
+        max_timeout_seconds=max_timeout_seconds,
+    )
+    # update agent registry
+    if custom_agents is not None:
+        for agent_class in custom_agents:
+            AgentBase.register_agent_class(agent_class=agent_class)
+
+    async def shutdown_signal_handler() -> None:
+        logger.info(
+            f"Received shutdown signal. Gracefully stopping the server at "
+            f"[{host}:{port}].",
+        )
+        await server.stop(grace=5)
+
+    loop = asyncio.get_running_loop()
+    if os.name != "nt":
+        # windows does not support add_signal_handler
+        for sig in (signal.SIGINT, signal.SIGTERM):
+            loop.add_signal_handler(
+                sig,
+                lambda: asyncio.create_task(shutdown_signal_handler()),
+            )
+    while True:
+        try:
+            port = check_port(port)
+            servicer.port = port
+            server = grpc.aio.server(
+                futures.ThreadPoolExecutor(max_workers=None),
+            )
+            add_RpcAgentServicer_to_server(servicer, server)
+            if local_mode:
+                server.add_insecure_port(f"localhost:{port}")
+            else:
+                server.add_insecure_port(f"0.0.0.0:{port}")
+            await server.start()
+            break
+        except OSError:
+            logger.warning(
+                f"Failed to start agent server at port [{port}]"
+                f"try another port",
+            )
+    logger.info(
+        f"agent server [{server_id}] at {host}:{port} started successfully",
+    )
+    if start_event is not None:
+        pipe.send(port)
+        start_event.set()
+        while not stop_event.is_set():
+            await asyncio.sleep(1)
+        logger.info(
+            f"Stopping agent server at [{host}:{port}]",
+        )
+        await server.stop(grace=10.0)
+    else:
+        await server.wait_for_termination()
+    logger.info(
+        f"agent server [{server_id}] at {host}:{port} stopped successfully",
+    )
+
+
+class RpcAgentServerLauncher:
+    """The launcher of AgentServer."""
+
+    def __init__(
+        self,
+        host: str = "localhost",
+        port: int = None,
+        max_pool_size: int = 8192,
+        max_timeout_seconds: int = 1800,
+        local_mode: bool = False,
+        custom_agents: list = None,
+        server_id: str = None,
+        agent_class: Type[AgentBase] = None,
+        agent_args: tuple = (),
+        agent_kwargs: dict = None,
+    ) -> None:
+        """Init a launcher of agent server.
+
+        Args:
+            host (`str`, defaults to `"localhost"`):
+                Hostname of the agent server.
+            port (`int`, defaults to `None`):
+                Socket port of the agent server.
+            max_pool_size (`int`, defaults to `8192`):
+                The max number of agent reply messages that the server can
+                accommodate. Note that the oldest message will be deleted
+                after exceeding the pool size.
+            max_timeout_seconds (`int`, defaults to `1800`):
+                Maximum time for reply messages to be cached in the server.
+                Note that expired messages will be deleted.
+            local_mode (`bool`, defaults to `False`):
+                If `True`, only listen to requests from "localhost", otherwise,
+                listen to requests from all hosts.
+            custom_agents (`list`, defaults to `None`):
+                A list of custom agent classes that are not in
+                `agentscope.agents`.
+            server_id (`str`, defaults to `None`):
+                The id of the agent server. If not specified, a random id
+                will be generated.
+            agent_class (`Type[AgentBase]`, deprecated):
+                The AgentBase subclass encapsulated by this wrapper.
+            agent_args (`tuple`, deprecated): The args tuple used to
+                initialize the agent_class.
+            agent_kwargs (`dict`, deprecated): The args dict used to
+                initialize the agent_class.
+        """
+        self.host = host
+        self.port = check_port(port)
+        self.max_pool_size = max_pool_size
+        self.max_timeout_seconds = max_timeout_seconds
+        self.local_mode = local_mode
+        self.server = None
+        self.stop_event = None
+        self.parent_con = None
+        self.custom_agents = custom_agents
+        self.server_id = (
+            self.generate_server_id() if server_id is None else server_id
+        )
+        if (
+            agent_class is not None
+            or len(agent_args) > 0
+            or agent_kwargs is not None
+        ):
+            logger.warning(
+                "`agent_class`, `agent_args` and `agent_kwargs` is deprecated"
+                " in `RpcAgentServerLauncher`",
+            )
+
+    def generate_server_id(self) -> str:
+        """Generate server id"""
+        return f"{self.host}:{self.port}-{_get_timestamp('%y%m%d-%H:%M:%S')}"
+
+    def _launch_in_main(self) -> None:
+        """Launch agent server in main-process"""
+        logger.info(
+            f"Launching agent server at [{self.host}:{self.port}]...",
+        )
+        asyncio.run(
+            _setup_agent_server_async(
+                host=self.host,
+                port=self.port,
+                server_id=self.server_id,
+                max_pool_size=self.max_pool_size,
+                max_timeout_seconds=self.max_timeout_seconds,
+                local_mode=self.local_mode,
+                custom_agents=self.custom_agents,
+            ),
+        )
+
+    def _launch_in_sub(self) -> None:
+        """Launch an agent server in sub-process."""
+        from agentscope._init import _INIT_SETTINGS
+
+        self.stop_event = Event()
+        self.parent_con, child_con = Pipe()
+        start_event = Event()
+        server_process = Process(
+            target=_setup_agent_server,
+            kwargs={
+                "host": self.host,
+                "port": self.port,
+                "server_id": self.server_id,
+                "init_settings": _INIT_SETTINGS,
+                "start_event": start_event,
+                "stop_event": self.stop_event,
+                "pipe": child_con,
+                "max_pool_size": self.max_pool_size,
+                "max_timeout_seconds": self.max_timeout_seconds,
+                "local_mode": self.local_mode,
+                "custom_agents": self.custom_agents,
+            },
+        )
+        server_process.start()
+        self.port = self.parent_con.recv()
+        start_event.wait()
+        self.server = server_process
+        logger.info(
+            f"Launch agent server at [{self.host}:{self.port}] success",
+        )
+
+    def launch(self, in_subprocess: bool = True) -> None:
+        """launch an agent server.
+
+        Args:
+            in_subprocess (bool, optional): launch the server in subprocess.
+                Defaults to True. For agents that need to obtain command line
+                input, such as UserAgent, please set this value to False.
+        """
+        if in_subprocess:
+            self._launch_in_sub()
+        else:
+            self._launch_in_main()
+
+    def wait_until_terminate(self) -> None:
+        """Wait for server process"""
+        if self.server is not None:
+            self.server.join()
+
+    def shutdown(self) -> None:
+        """Shutdown the agent server."""
+        if self.server is not None:
+            if self.stop_event is not None:
+                self.stop_event.set()
+                self.stop_event = None
+            self.server.join()
+            if self.server.is_alive():
+                self.server.kill()
+                logger.info(
+                    f"Agent server at port [{self.port}] is killed.",
+                )
+            self.server = None
+
+
+def as_server() -> None:
+    """Launch an agent server with terminal command.
+
+    Note:
+
+        The arguments of `as_server` are listed as follows:
+
+        * `--host`: the hostname of the server.
+        * `--port`: the socket port of the server.
+        * `--max-pool-size`: max number of agent reply messages that the server
+          can accommodate. Note that the oldest message will be deleted
+          after exceeding the pool size.
+        * `--max-timeout-seconds`: max time for reply messages to be cached
+          in the server. Note that expired messages will be deleted.
+        * `--local-mode`: whether the started agent server only listens to
+          local requests.
+        * `--model-config-path`: the path to the model config json file
+
+        In most cases, you only need to specify the `--host`, `--port` and
+        `--model-config-path`.
+
+        .. code-block:: shell
+
+            as_server --host localhost --port 12345 --model-config-path config.json
+
+    """  # noqa
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--host",
+        type=str,
+        default="localhost",
+        help="hostname of the server",
+    )
+    parser.add_argument(
+        "--port",
+        type=int,
+        default=12310,
+        help="socket port of the server",
+    )
+    parser.add_argument(
+        "--max-pool-size",
+        type=int,
+        default=8192,
+        help=(
+            "max number of agent reply messages that the server "
+            "can accommodate. Note that the oldest message will be deleted "
+            "after exceeding the pool size."
+        ),
+    )
+    parser.add_argument(
+        "--max-timeout-seconds",
+        type=int,
+        default=1800,
+        help=(
+            "max time for agent reply messages to be cached"
+            "in the server. Note that expired messages will be deleted."
+        ),
+    )
+    parser.add_argument(
+        "--local-mode",
+        type=bool,
+        default=False,
+        help=(
+            "If `True`, only listen to requests from 'localhost', otherwise, "
+            "listen to requests from all hosts."
+        ),
+    )
+    parser.add_argument(
+        "--model-config-path",
+        type=str,
+        help="path to the model config json file",
+    )
+    args = parser.parse_args()
+    agentscope.init(
+        project="agent_server",
+        name=f"server_{args.host}:{args.port}",
+        runtime_id=_get_timestamp(
+            "server_{}_{}_%y%m%d-%H%M%S",
+        ).format(args.host, args.port),
+        model_configs=args.model_config_path,
+    )
+    launcher = RpcAgentServerLauncher(
+        host=args.host,
+        port=args.port,
+        max_pool_size=args.max_pool_size,
+        max_timeout_seconds=args.max_timeout_seconds,
+        local_mode=args.local_mode,
+    )
+    launcher.launch(in_subprocess=False)
+    launcher.wait_until_terminate()
diff --git a/src/agentscope/server/servicer.py b/src/agentscope/server/servicer.py
new file mode 100644
index 000000000..53c63425f
--- /dev/null
+++ b/src/agentscope/server/servicer.py
@@ -0,0 +1,313 @@
+# -*- coding: utf-8 -*-
+""" Server of distributed agent"""
+import threading
+import base64
+import json
+import traceback
+from concurrent import futures
+from loguru import logger
+
+try:
+    import dill
+    import grpc
+    from grpc import ServicerContext
+    from expiringdict import ExpiringDict
+    from ..rpc.rpc_agent_pb2 import RpcMsg  # pylint: disable=E0611
+    from ..rpc.rpc_agent_pb2_grpc import RpcAgentServicer
+except ImportError as import_error:
+    from agentscope.utils.tools import ImportErrorReporter
+
+    dill = ImportErrorReporter(import_error, "distribute")
+    grpc = ImportErrorReporter(import_error, "distribute")
+    ServicerContext = ImportErrorReporter(import_error, "distribute")
+    ExpiringDict = ImportErrorReporter(import_error, "distribute")
+    RpcMsg = ImportErrorReporter(  # type: ignore[misc]
+        import_error,
+        "distribute",
+    )
+    RpcAgentServicer = ImportErrorReporter(import_error, "distribute")
+
+from ..agents.agent import AgentBase
+from ..message import (
+    Msg,
+    PlaceholderMessage,
+    deserialize,
+)
+
+
+class AgentServerServicer(RpcAgentServicer):
+    """A Servicer for RPC Agent Server (formerly RpcServerSideWrapper)"""
+
+    def __init__(
+        self,
+        host: str = "localhost",
+        port: int = None,
+        max_pool_size: int = 8192,
+        max_timeout_seconds: int = 1800,
+    ):
+        """Init the AgentServerServicer.
+
+        Args:
+            host (`str`, defaults to "localhost"):
+                Hostname of the rpc agent server.
+            port (`int`, defaults to `None`):
+                Port of the rpc agent server.
+            max_pool_size (`int`, defaults to `8192`):
+                The max number of agent reply messages that the server can
+                accommodate. Note that the oldest message will be deleted
+                after exceeding the pool size.
+            max_timeout_seconds (`int`, defaults to `1800`):
+                Maximum time for reply messages to be cached in the server.
+                Note that expired messages will be deleted.
+        """
+        self.host = host
+        self.port = port
+        self.result_pool = ExpiringDict(
+            max_len=max_pool_size,
+            max_age_seconds=max_timeout_seconds,
+        )
+        self.executor = futures.ThreadPoolExecutor(max_workers=None)
+        self.task_id_lock = threading.Lock()
+        self.agent_id_lock = threading.Lock()
+        self.task_id_counter = 0
+        self.agent_pool: dict[str, AgentBase] = {}
+
+    def get_task_id(self) -> int:
+        """Get the auto-increment task id.
+        Each reply call will get a unique task id."""
+        with self.task_id_lock:
+            self.task_id_counter += 1
+            return self.task_id_counter
+
+    def agent_exists(self, agent_id: str) -> bool:
+        """Check whether the agent exists.
+
+        Args:
+            agent_id (`str`): the agent id.
+
+        Returns:
+            bool: whether the agent exists.
+        """
+        return agent_id in self.agent_pool
+
+    def check_and_generate_agent(
+        self,
+        agent_id: str,
+        agent_configs: dict,
+    ) -> None:
+        """
+        Check whether the agent exists, and create new agent instance
+        for new agent.
+
+        Args:
+            agent_id (`str`): the agent id.
+            agent_configs (`dict`): configuration used to initialize the agent,
+                with three fields (generated in `_AgentMeta`):
+
+                .. code-block:: python
+
+                    {
+                        "class_name": {name of the agent}
+                        "args": {args in tuple type to init the agent}
+                        "kwargs": {args in dict type to init the agent}
+                    }
+
+        """
+        with self.agent_id_lock:
+            if agent_id not in self.agent_pool:
+                agent_class_name = agent_configs["class_name"]
+                agent_instance = AgentBase.get_agent_class(agent_class_name)(
+                    *agent_configs["args"],
+                    **agent_configs["kwargs"],
+                )
+                agent_instance._agent_id = agent_id  # pylint: disable=W0212
+                self.agent_pool[agent_id] = agent_instance
+                logger.info(f"create agent instance [{agent_id}]")
+
+    def check_and_delete_agent(self, agent_id: str) -> None:
+        """
+        Check whether the agent exists, and delete the agent instance
+        for the agent_id.
+
+        Args:
+            agent_id (`str`): the agent id.
+        """
+        with self.agent_id_lock:
+            if agent_id in self.agent_pool:
+                self.agent_pool.pop(agent_id)
+                logger.info(f"delete agent instance [{agent_id}]")
+
+    def call_func(  # pylint: disable=W0236
+        self,
+        request: RpcMsg,
+        context: ServicerContext,
+    ) -> RpcMsg:
+        """Call the specific servicer function."""
+        if hasattr(self, request.target_func):
+            if request.target_func not in ["_create_agent", "_get"]:
+                if not self.agent_exists(request.agent_id):
+                    return context.abort(
+                        grpc.StatusCode.INVALID_ARGUMENT,
+                        f"Agent [{request.agent_id}] not exists.",
+                    )
+            return getattr(self, request.target_func)(request)
+        else:
+            # TODO: support other user defined method
+            logger.error(f"Unsupported method {request.target_func}")
+            return context.abort(
+                grpc.StatusCode.INVALID_ARGUMENT,
+                f"Unsupported method {request.target_func}",
+            )
+
+    def _reply(self, request: RpcMsg) -> RpcMsg:
+        """Call function of RpcAgentService
+
+        Args:
+            request (`RpcMsg`):
+                Message containing input parameters or input parameter
+                placeholders.
+
+        Returns:
+            `RpcMsg`: A serialized Msg instance with attributes name, host,
+            port and task_id
+        """
+        if request.value:
+            msg = deserialize(request.value)
+        else:
+            msg = None
+        task_id = self.get_task_id()
+        self.result_pool[task_id] = threading.Condition()
+        self.executor.submit(
+            self.process_messages,
+            task_id,
+            request.agent_id,
+            msg,  # type: ignore[arg-type]
+        )
+        return RpcMsg(
+            value=Msg(  # type: ignore[arg-type]
+                name=self.agent_pool[request.agent_id].name,
+                content=None,
+                task_id=task_id,
+            ).serialize(),
+        )
+
+    def _get(self, request: RpcMsg) -> RpcMsg:
+        """Get a reply message with specific task_id.
+
+        Args:
+            request (`RpcMsg`):
+                The task id that generated this message, with json format::
+
+                {
+                    'task_id': int
+                }
+
+        Returns:
+            `RpcMsg`: Concrete values of the specific message (or part of it).
+        """
+        msg = json.loads(request.value)
+        while True:
+            result = self.result_pool.get(msg["task_id"])
+            if isinstance(result, threading.Condition):
+                with result:
+                    result.wait(timeout=1)
+            else:
+                break
+        return RpcMsg(value=result.serialize())
+
+    def _observe(self, request: RpcMsg) -> RpcMsg:
+        """Observe function of the original agent.
+
+        Args:
+            request (`RpcMsg`):
+                The serialized input to be observed.
+
+        Returns:
+            `RpcMsg`: Empty RpcMsg.
+        """
+        msgs = deserialize(request.value)
+        for msg in msgs:
+            if isinstance(msg, PlaceholderMessage):
+                msg.update_value()
+        self.agent_pool[request.agent_id].observe(msgs)
+        return RpcMsg()
+
+    def _create_agent(self, request: RpcMsg) -> RpcMsg:
+        """Create a new agent instance with the given agent_id.
+
+        Args:
+            request (RpcMsg): request message with a `agent_id` field.
+        """
+        self.check_and_generate_agent(
+            request.agent_id,
+            agent_configs=(
+                dill.loads(base64.b64decode(request.value))
+                if request.value
+                else None
+            ),
+        )
+        return RpcMsg()
+
+    def _clone_agent(self, request: RpcMsg) -> RpcMsg:
+        """Clone a new agent instance from the origin instance.
+
+        Args:
+            request (RpcMsg): The `agent_id` field is the agent_id of the
+            agent to be cloned.
+
+        Returns:
+            `RpcMsg`: The `value` field contains the agent_id of generated
+            agent.
+        """
+        agent_id = request.agent_id
+        with self.agent_id_lock:
+            if agent_id not in self.agent_pool:
+                raise ValueError(f"Agent [{agent_id}] not exists")
+            ori_agent = self.agent_pool[agent_id]
+        new_agent = ori_agent.__class__(
+            *ori_agent._init_settings["args"],  # pylint: disable=W0212
+            **ori_agent._init_settings["kwargs"],  # pylint: disable=W0212
+        )
+        with self.agent_id_lock:
+            self.agent_pool[new_agent.agent_id] = new_agent
+        return RpcMsg(value=new_agent.agent_id)  # type: ignore[arg-type]
+
+    def _delete_agent(self, request: RpcMsg) -> RpcMsg:
+        """Delete the agent instance of the specific agent_id.
+
+        Args:
+            request (RpcMsg): request message with a `agent_id` field.
+        """
+        self.check_and_delete_agent(request.agent_id)
+        return RpcMsg()
+
+    def process_messages(
+        self,
+        task_id: int,
+        agent_id: str,
+        task_msg: dict = None,
+    ) -> None:
+        """Processing an input message and generate its reply message.
+
+        Args:
+            task_id (`int`): task id of the input message, .
+            agent_id (`str`): the id of the agent that accepted the message.
+            task_msg (`dict`): the input message.
+        """
+        if isinstance(task_msg, PlaceholderMessage):
+            task_msg.update_value()
+        cond = self.result_pool[task_id]
+        try:
+            result = self.agent_pool[agent_id].reply(task_msg)
+            self.result_pool[task_id] = result
+        except Exception:
+            error_msg = traceback.format_exc()
+            logger.error(f"Error in agent [{agent_id}]:\n{error_msg}")
+            self.result_pool[task_id] = Msg(
+                name="ERROR",
+                role="assistant",
+                __status="ERROR",
+                content=f"Error in agent [{agent_id}]:\n{error_msg}",
+            )
+        with cond:
+            cond.notify_all()

From 5b6cf2cca2a4ee865170670a3eeecb0475810e6a Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Fri, 24 May 2024 12:14:26 +0800
Subject: [PATCH 31/55] Delete examples/distributed_simulation/run_simlation.sh

---
 .../distributed_simulation/run_simlation.sh   | 25 -------------------
 1 file changed, 25 deletions(-)
 delete mode 100644 examples/distributed_simulation/run_simlation.sh

diff --git a/examples/distributed_simulation/run_simlation.sh b/examples/distributed_simulation/run_simlation.sh
deleted file mode 100644
index 6fac7c4c4..000000000
--- a/examples/distributed_simulation/run_simlation.sh
+++ /dev/null
@@ -1,25 +0,0 @@
-#!/bin/bash
-
-# default values
-base_port=12330
-hosts="localhost" # or "server1 server2 server3 ..."
-moderator_per_host=4
-model_per_host=8
-agent_type="random" # or "llm"
-max_value=100
-
-# check server-per-host
-if ! [[ "$1" =~ ^[0-9]+$ ]]; then
-    echo "Usage: $0 <server-per-host> <participant-num>"
-    exit 1
-fi
-
-# check participant-num
-if ! [[ "$2" =~ ^[0-9]+$ ]]; then
-    echo "Usage: $0 <server-per-host> <participant-num>"
-    exit 1
-fi
-
-mkdir -p log
-
-python main.py --role main --hosts ${hosts} --base-port ${base_port} --participant-num $2 --server-per-host $1 --model-per-host ${model_per_host} --moderator-per-host ${moderator_per_host} --agent-type ${agent_type} --max-value ${max_value}

From 28037bc76b993bce6aab7a553dcd74eacd4dd753 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Sat, 25 May 2024 19:38:39 +0800
Subject: [PATCH 32/55] renamed `Finetune_DialogAgent` to `FinetuneDialogAgent`

---
 .../conversation_with_agent_with_finetuned_model/README.md  | 2 +-
 .../conversation_with_agent_with_finetuned_model.py         | 6 +++---
 .../finetune_dialogagent.py                                 | 6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
index d9f682a5f..06f405c44 100644
--- a/examples/conversation_with_agent_with_finetuned_model/README.md
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -9,7 +9,7 @@ Compared to basic conversation setup, this example introduces model loading and
 - Initialize an agent or use `dialog_agent.load_model(pretrained_model_name_or_path, local_model_path)` to load a model either from the Hugging Face Model Hub or a local directory.
 - Initalize an agent or apply `dialog_agent.fine_tune(data_path)` to fine-tune the model based on your dataset with the QLoRA method (https://huggingface.co/blog/4bit-transformers-bitsandbytes).
 
-The default hyperparameters for (SFT) fine-tuning are specified in `agentscope/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py` and `agentscope/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json`. For customized hyperparameters, specify them in `model_configs` if the model needs to be fine-tuned at initialization, or specify through `fine_tune_config` in `Finetune_DialogAgent`'s `fine_tune` method after initialization, as shown in the example script `conversation_with_agent_with_finetuned_model.py`.
+The default hyperparameters for (SFT) fine-tuning are specified in `agentscope/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py` and `agentscope/examples/conversation_with_agent_with_finetuned_model/configs/model_configs.json`. For customized hyperparameters, specify them in `model_configs` if the model needs to be fine-tuned at initialization, or specify through `fine_tune_config` in `FinetuneDialogAgent`'s `fine_tune` method after initialization, as shown in the example script `conversation_with_agent_with_finetuned_model.py`.
 
 ## Agent Initialization
 
diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 244b7ec5d..a2c2190e4 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -2,7 +2,7 @@
 """
 This script sets up a conversational agent using
 AgentScope with a Hugging Face model.
-It includes initializing a Finetune_DialogAgent,
+It includes initializing a FinetuneDialogAgent,
 loading and fine-tuning a pre-trained model,
 and conducting a dialogue via a sequential pipeline.
 The conversation continues until the user exits.
@@ -11,7 +11,7 @@
 """
 # pylint: disable=unused-import
 from huggingface_model import HuggingFaceWrapper
-from finetune_dialogagent import Finetune_DialogAgent
+from FinetuneDialogAgent import FinetuneDialogAgent
 import agentscope
 from agentscope.agents.user_agent import UserAgent
 from agentscope.pipelines.functional import sequentialpipeline
@@ -66,7 +66,7 @@ def main() -> None:
     # )
 
     # Init agents with the custom model
-    dialog_agent = Finetune_DialogAgent(
+    dialog_agent = FinetuneDialogAgent(
         name="Assistant",
         sys_prompt=(
             "Explain in simple terms how the attention mechanism of "
diff --git a/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py b/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
index 60a68ca31..b708ca81b 100644
--- a/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
+++ b/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
@@ -1,6 +1,6 @@
 # -*- coding: utf-8 -*-
 """
-This module provides the Finetune_DialogAgent class,
+This module provides the FinetuneDialogAgent class,
 which extends DialogAgent to enhance fine-tuning
 capabilities with custom hyperparameters.
 """
@@ -11,7 +11,7 @@
 from agentscope.agents import DialogAgent
 
 
-class Finetune_DialogAgent(DialogAgent):
+class FinetuneDialogAgent(DialogAgent):
     """
     A dialog agent capable of fine-tuning its
     underlying model based on provided data.
@@ -29,7 +29,7 @@ def __init__(
         memory_config: Optional[dict] = None,
     ):
         """
-        Initializes a new Finetune_DialogAgent with specified configuration.
+        Initializes a new FinetuneDialogAgent with specified configuration.
 
         Arguments:
             name (str): Name of the agent.

From 3fcc03b18c24d271e2a938244e0977e17e515d9b Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Sat, 25 May 2024 21:05:49 +0800
Subject: [PATCH 33/55] rename `finetune_dialogAgent` to `FinetuneDialogAgent`

---
 .../{finetune_dialogagent.py => FinetuneDialogAgent.py}           | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 rename examples/conversation_with_agent_with_finetuned_model/{finetune_dialogagent.py => FinetuneDialogAgent.py} (100%)

diff --git a/examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
similarity index 100%
rename from examples/conversation_with_agent_with_finetuned_model/finetune_dialogagent.py
rename to examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py

From 3fb3540b65d279fe68686696977a88c6a826f159 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Tue, 28 May 2024 22:52:38 +0800
Subject: [PATCH 34/55] Updated README to be more precise in its description

---
 .../conversation_with_agent_with_finetuned_model/README.md    | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
index 06f405c44..847888a0e 100644
--- a/examples/conversation_with_agent_with_finetuned_model/README.md
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -1,6 +1,6 @@
-# Multi-Agent Conversation with Custom Model Loading and Fine-Tuning in AgentScope
+# User-Agent Conversation with Custom Model Loading and Fine-Tuning in AgentScope
 
-This example demonstrates how to load and optionally fine-tune a Hugging Face model within a multi-agent conversation setup using AgentScope. The complete code is provided in `agentscope/examples/conversation_with_agent_with_finetuned_model`.
+This example demonstrates how to load and optionally fine-tune a Hugging Face model within a user-agent conversation setup using AgentScope. The complete code is provided in `agentscope/examples/conversation_with_agent_with_finetuned_model`.
 
 ## Functionality Overview
 

From ce1373bba6a6b562578cde83665a62e9c6af19b4 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 30 May 2024 17:16:28 +0800
Subject: [PATCH 35/55] fixed error on quantization/QLora

---
 .../conversation_with_agent_with_finetuned_model.py             | 2 +-
 .../huggingface_model.py                                        | 1 +
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index a2c2190e4..233dbfee1 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -53,7 +53,7 @@ def main() -> None:
                         "load_in_4bit": True,
                         "bnb_4bit_use_double_quant": True,
                         "bnb_4bit_quant_type": "nf4",
-                        "bnb_4bit_compute_dtype": "torch.bfloat16",
+                        "bnb_4bit_compute_dtype": "bfloat16",
                     },
                 },
             },
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 809c1fd1a..8f4604638 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -95,6 +95,7 @@ def __init__(
         self.load_model(
             pretrained_model_name_or_path,
             local_model_path=local_model_path,
+            fine_tune_config=fine_tune_config,
         )
         self.load_tokenizer(
             pretrained_model_name_or_path,

From b2f2b09cbaa1348d7b78039ce7d2b8bc7aa85a2d Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 30 May 2024 18:55:55 +0800
Subject: [PATCH 36/55] add required dependencies for some use cases

---
 examples/conversation_with_agent_with_finetuned_model/README.md | 1 +
 1 file changed, 1 insertion(+)

diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
index 847888a0e..99e3fcaa7 100644
--- a/examples/conversation_with_agent_with_finetuned_model/README.md
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -67,6 +67,7 @@ Before running this example, ensure you have installed the following packages:
 - `datasets`
 - `trl`
 - `bitsandbytes`
+- `sentencepiece`
 
 Additionally, set `HUGGINGFACE_TOKEN` in the `agentscope/examples/conversation_with_agent_with_finetuned_model/.env`.
 

From 1d0704d544e80d30ecd5d12db35a64c73f0a60af Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Thu, 30 May 2024 19:39:53 +0800
Subject: [PATCH 37/55] optimized the behavior of `device_map` when loading a
 huggingface model.

optimized the behavior of `device_map` when loading a huggingface model. Now if `device` is not given by the user, `device_map="auto"` by default; otherwise  `device_map` is set to the user-specified `device`.
---
 .../huggingface_model.py                                    | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 8f4604638..97315ff8e 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -86,10 +86,12 @@ def __init__(
         self.huggingface_token = os.getenv("HUGGINGFACE_TOKEN")
 
         if device is None:
+            self.device_map = 'auto'
             self.device = torch.device(
                 "cuda" if torch.cuda.is_available() else "cpu",
             )
         else:
+            self.device_map = device
             self.device = device
 
         self.load_model(
@@ -251,7 +253,7 @@ def load_model(
                     pretrained_model_name_or_path,
                     quantization_config=bnb_config,
                     token=self.huggingface_token,
-                    device_map="auto",
+                    device_map=self.device_map,
                 )
                 info_msg = (
                     f"Successfully loaded new model "
@@ -263,7 +265,7 @@ def load_model(
                     local_model_path,
                     quantization_config=bnb_config,
                     local_files_only=True,
-                    device_map="auto",
+                    device_map=self.device_map,
                 )
                 info_msg = (
                     f"Successfully loaded new model "

From 86852139d8107fcebf974ccecbcfd66d67eaec94 Mon Sep 17 00:00:00 2001
From: zyzhang <zyzhang1130@gmail.com>
Date: Thu, 30 May 2024 13:04:22 +0000
Subject: [PATCH 38/55] optimized the behavior of device_map when loading a
 huggingface model.

---
 .../huggingface_model.py                                        | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 97315ff8e..4502b5ad5 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -86,7 +86,7 @@ def __init__(
         self.huggingface_token = os.getenv("HUGGINGFACE_TOKEN")
 
         if device is None:
-            self.device_map = 'auto'
+            self.device_map = "auto"
             self.device = torch.device(
                 "cuda" if torch.cuda.is_available() else "cpu",
             )

From 3da24f446cd67d87ec81ded714330aceaf734569 Mon Sep 17 00:00:00 2001
From: zyzhang <zyzhang1130@gmail.com>
Date: Thu, 30 May 2024 13:42:46 +0000
Subject: [PATCH 39/55] optimized the behavior of  when loading a huggingface
 model (default to disable ).

---
 .../huggingface_model.py                          | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 4502b5ad5..ea4b8cd90 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -240,20 +240,22 @@ def load_model(
         """
 
         bnb_config_default = {}
+        bnb_config = {}
 
         if fine_tune_config is not None:
             if fine_tune_config.get("bnb_config") is not None:
                 bnb_config_default.update(fine_tune_config["bnb_config"])
-
-        bnb_config = BitsAndBytesConfig(**bnb_config_default)
+        if bnb_config != bnb_config_default:
+            bnb_config = BitsAndBytesConfig(**bnb_config_default)
 
         try:
             if local_model_path is None:
                 self.model = AutoModelForCausalLM.from_pretrained(
                     pretrained_model_name_or_path,
-                    quantization_config=bnb_config,
-                    token=self.huggingface_token,
                     device_map=self.device_map,
+                    torch_dtype=torch.bfloat16,
+                    **({"quantization_config": bnb_config} if bnb_config != {} else {}),
+                    token = self.huggingface_token,
                 )
                 info_msg = (
                     f"Successfully loaded new model "
@@ -263,9 +265,10 @@ def load_model(
             else:
                 self.model = AutoModelForCausalLM.from_pretrained(
                     local_model_path,
-                    quantization_config=bnb_config,
-                    local_files_only=True,
                     device_map=self.device_map,
+                    torch_dtype=torch.bfloat16,
+                    **({"quantization_config": bnb_config} if bnb_config != {} else {}),
+                    local_files_only=True,
                 )
                 info_msg = (
                     f"Successfully loaded new model "

From 2ea87031d6b8d41e4aa663d4fbd72d55ade6c4fe Mon Sep 17 00:00:00 2001
From: zyzhang <zyzhang1130@gmail.com>
Date: Thu, 30 May 2024 13:56:54 +0000
Subject: [PATCH 40/55] optimized the behavior of  when loading a huggingface
 model (default to disable ).

---
 .../huggingface_model.py                           | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index ea4b8cd90..4e5084552 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -254,8 +254,12 @@ def load_model(
                     pretrained_model_name_or_path,
                     device_map=self.device_map,
                     torch_dtype=torch.bfloat16,
-                    **({"quantization_config": bnb_config} if bnb_config != {} else {}),
-                    token = self.huggingface_token,
+                    **(
+                        {"quantization_config": bnb_config}
+                        if bnb_config != {}
+                        else {}
+                    ),
+                    token=self.huggingface_token,
                 )
                 info_msg = (
                     f"Successfully loaded new model "
@@ -267,7 +271,11 @@ def load_model(
                     local_model_path,
                     device_map=self.device_map,
                     torch_dtype=torch.bfloat16,
-                    **({"quantization_config": bnb_config} if bnb_config != {} else {}),
+                    **(
+                        {"quantization_config": bnb_config}
+                        if bnb_config != {}
+                        else {}
+                    ),
                     local_files_only=True,
                 )
                 info_msg = (

From e0ba28284036f468137cd83481e8dad8600dd2df Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Fri, 7 Jun 2024 05:07:32 +0800
Subject: [PATCH 41/55] updated peft config

now the user can choose to do full-parameter finetuning by not passing `lora_config`
---
 ...rsation_with_agent_with_finetuned_model.py |  2 +-
 .../huggingface_model.py                      | 29 ++++++++++---------
 2 files changed, 16 insertions(+), 15 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 233dbfee1..f915a3d2a 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -48,7 +48,7 @@ def main() -> None:
                 # the standard lora and sfttrainer fields.
                 "fine_tune_config": {
                     "lora_config": {"r": 16, "lora_alpha": 32},
-                    "training_args": {"max_steps": 200, "logging_steps": 1},
+                    "training_args": {"num_train_epochs": 5, "logging_steps": 1},
                     "bnb_config": {
                         "load_in_4bit": True,
                         "bnb_4bit_use_double_quant": True,
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 4e5084552..699a21359 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -239,13 +239,13 @@ def load_model(
                        or network issues while fetching the model.
         """
 
+        bnb_config = None
         bnb_config_default = {}
-        bnb_config = {}
 
         if fine_tune_config is not None:
             if fine_tune_config.get("bnb_config") is not None:
                 bnb_config_default.update(fine_tune_config["bnb_config"])
-        if bnb_config != bnb_config_default:
+        if bnb_config_default != {}:
             bnb_config = BitsAndBytesConfig(**bnb_config_default)
 
         try:
@@ -256,7 +256,7 @@ def load_model(
                     torch_dtype=torch.bfloat16,
                     **(
                         {"quantization_config": bnb_config}
-                        if bnb_config != {}
+                        if bnb_config is not None
                         else {}
                     ),
                     token=self.huggingface_token,
@@ -273,7 +273,7 @@ def load_model(
                     torch_dtype=torch.bfloat16,
                     **(
                         {"quantization_config": bnb_config}
-                        if bnb_config != {}
+                        if bnb_config is not None
                         else {}
                     ),
                     local_files_only=True,
@@ -344,6 +344,7 @@ def load_tokenizer(
                     f"'{pretrained_model_name_or_path}'"
                     f" from '{local_tokenizer_path}'",
                 )
+            self.tokenizer.add_special_tokens({'pad_token': '[PAD]'})
 
         except Exception as e:
             # Handle exceptions during model loading,
@@ -483,13 +484,8 @@ def fine_tune_training(
 
         from peft import LoraConfig
 
-        lora_config_default = {
-            "r": 16,
-            "lora_alpha": 32,
-            "lora_dropout": 0.05,
-            "bias": "none",
-            "task_type": "CAUSAL_LM",
-        }
+        lora_config = None
+        lora_config_default = {}
 
         if fine_tune_config is not None:
             if fine_tune_config.get("lora_config") is not None:
@@ -514,8 +510,9 @@ def fine_tune_training(
 
         from peft import get_peft_model
 
-        lora_config = LoraConfig(**lora_config_default)
-        model = get_peft_model(model, lora_config)
+        if lora_config_default != {}:
+            lora_config = LoraConfig(**lora_config_default)
+            model = get_peft_model(model, lora_config)
 
         collator = DataCollatorForCompletionOnlyLM(
             response_template=" ### Answer:",
@@ -530,7 +527,11 @@ def fine_tune_training(
             data_collator=collator,
             train_dataset=formatted_dataset["train"],
             eval_dataset=formatted_dataset["test"],
-            peft_config=lora_config,
+            **(
+                {"peft_config": lora_config}
+                if lora_config is not None
+                else {}
+            ),
             args=trainer_args,
             max_seq_length=2048,
         )

From 7f2f56537dcc48040a233f19df2532d11c7c8a3e Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Fri, 7 Jun 2024 05:45:23 +0800
Subject: [PATCH 42/55] updated peft config

now the user can choose to do full-parameter finetuning by not passing `lora_config` and `bnb_config`
---
 .../huggingface_model.py                      | 22 +++++++------------
 1 file changed, 8 insertions(+), 14 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 699a21359..6367382fe 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -18,6 +18,8 @@
     BitsAndBytesConfig,
 )
 import transformers
+from peft import LoraConfig
+from peft import get_peft_model
 from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
 from datasets import load_dataset
 from loguru import logger
@@ -245,7 +247,7 @@ def load_model(
         if fine_tune_config is not None:
             if fine_tune_config.get("bnb_config") is not None:
                 bnb_config_default.update(fine_tune_config["bnb_config"])
-        if bnb_config_default != {}:
+        if bnb_config_default:
             bnb_config = BitsAndBytesConfig(**bnb_config_default)
 
         try:
@@ -344,7 +346,7 @@ def load_tokenizer(
                     f"'{pretrained_model_name_or_path}'"
                     f" from '{local_tokenizer_path}'",
                 )
-            self.tokenizer.add_special_tokens({'pad_token': '[PAD]'})
+            self.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
 
         except Exception as e:
             # Handle exceptions during model loading,
@@ -482,15 +484,9 @@ def fine_tune_training(
 
         formatted_dataset = dataset_reduced.train_test_split(test_size=0.1)
 
-        from peft import LoraConfig
-
         lora_config = None
         lora_config_default = {}
 
-        if fine_tune_config is not None:
-            if fine_tune_config.get("lora_config") is not None:
-                lora_config_default.update(fine_tune_config["lora_config"])
-
         training_defaults = {
             "per_device_train_batch_size": 1,
             "gradient_accumulation_steps": 4,
@@ -502,15 +498,15 @@ def fine_tune_training(
         }
 
         if fine_tune_config is not None:
+            if fine_tune_config.get("lora_config") is not None:
+                lora_config_default.update(fine_tune_config["lora_config"])
             if fine_tune_config.get("training_args") is not None:
                 training_defaults.update(fine_tune_config["training_args"])
 
         if output_dir is not None:
             training_defaults["output_dir"] = output_dir
 
-        from peft import get_peft_model
-
-        if lora_config_default != {}:
+        if lora_config_default:
             lora_config = LoraConfig(**lora_config_default)
             model = get_peft_model(model, lora_config)
 
@@ -528,9 +524,7 @@ def fine_tune_training(
             train_dataset=formatted_dataset["train"],
             eval_dataset=formatted_dataset["test"],
             **(
-                {"peft_config": lora_config}
-                if lora_config is not None
-                else {}
+                {"peft_config": lora_config} if lora_config is not None else {}
             ),
             args=trainer_args,
             max_seq_length=2048,

From 427f9f51ad8dff2e1d3d0d3b4e97110335264aee Mon Sep 17 00:00:00 2001
From: Ze Yu Zhang <zyzhang1130@gmail.com>
Date: Fri, 7 Jun 2024 05:52:10 +0800
Subject: [PATCH 43/55] now the user can choose to do full-parameter finetuning
 by not passing  and

---
 ...rsation_with_agent_with_finetuned_model.py |  5 ++++-
 .../huggingface_model.py                      | 22 +++++++------------
 2 files changed, 12 insertions(+), 15 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index f915a3d2a..6301c39c8 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -48,7 +48,10 @@ def main() -> None:
                 # the standard lora and sfttrainer fields.
                 "fine_tune_config": {
                     "lora_config": {"r": 16, "lora_alpha": 32},
-                    "training_args": {"num_train_epochs": 5, "logging_steps": 1},
+                    "training_args": {
+                        "num_train_epochs": 5,
+                        "logging_steps": 1,
+                    },
                     "bnb_config": {
                         "load_in_4bit": True,
                         "bnb_4bit_use_double_quant": True,
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 699a21359..6367382fe 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -18,6 +18,8 @@
     BitsAndBytesConfig,
 )
 import transformers
+from peft import LoraConfig
+from peft import get_peft_model
 from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
 from datasets import load_dataset
 from loguru import logger
@@ -245,7 +247,7 @@ def load_model(
         if fine_tune_config is not None:
             if fine_tune_config.get("bnb_config") is not None:
                 bnb_config_default.update(fine_tune_config["bnb_config"])
-        if bnb_config_default != {}:
+        if bnb_config_default:
             bnb_config = BitsAndBytesConfig(**bnb_config_default)
 
         try:
@@ -344,7 +346,7 @@ def load_tokenizer(
                     f"'{pretrained_model_name_or_path}'"
                     f" from '{local_tokenizer_path}'",
                 )
-            self.tokenizer.add_special_tokens({'pad_token': '[PAD]'})
+            self.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
 
         except Exception as e:
             # Handle exceptions during model loading,
@@ -482,15 +484,9 @@ def fine_tune_training(
 
         formatted_dataset = dataset_reduced.train_test_split(test_size=0.1)
 
-        from peft import LoraConfig
-
         lora_config = None
         lora_config_default = {}
 
-        if fine_tune_config is not None:
-            if fine_tune_config.get("lora_config") is not None:
-                lora_config_default.update(fine_tune_config["lora_config"])
-
         training_defaults = {
             "per_device_train_batch_size": 1,
             "gradient_accumulation_steps": 4,
@@ -502,15 +498,15 @@ def fine_tune_training(
         }
 
         if fine_tune_config is not None:
+            if fine_tune_config.get("lora_config") is not None:
+                lora_config_default.update(fine_tune_config["lora_config"])
             if fine_tune_config.get("training_args") is not None:
                 training_defaults.update(fine_tune_config["training_args"])
 
         if output_dir is not None:
             training_defaults["output_dir"] = output_dir
 
-        from peft import get_peft_model
-
-        if lora_config_default != {}:
+        if lora_config_default:
             lora_config = LoraConfig(**lora_config_default)
             model = get_peft_model(model, lora_config)
 
@@ -528,9 +524,7 @@ def fine_tune_training(
             train_dataset=formatted_dataset["train"],
             eval_dataset=formatted_dataset["test"],
             **(
-                {"peft_config": lora_config}
-                if lora_config is not None
-                else {}
+                {"peft_config": lora_config} if lora_config is not None else {}
             ),
             args=trainer_args,
             max_seq_length=2048,

From 432bb210e73359e20489f2f2b2524b9bb6be6eef Mon Sep 17 00:00:00 2001
From: Ze Yu Zhang <zyzhang1130@gmail.com>
Date: Fri, 7 Jun 2024 18:16:19 +0800
Subject: [PATCH 44/55] moved peft loading to ; removed independently saving
 tokenizer

---
 ...rsation_with_agent_with_finetuned_model.py | 12 +++-
 .../huggingface_model.py                      | 67 +++++++++----------
 2 files changed, 40 insertions(+), 39 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 6301c39c8..79ba887a9 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -30,8 +30,8 @@ def main() -> None:
                 # Or another generative model of your choice.
                 # Needed from loading from Hugging Face.
                 "pretrained_model_name_or_path": "google/gemma-7b",
-                # "local_model_path":  # Specify your local model path
-                # "local_tokenizer_path":  # Specify your local tokenizer path
+                # "local_model_path": , # Specify your local model path
+                # "local_tokenizer_path":,  # Specify your local tokenizer path
                 "max_length": 128,
                 # Device for inference. Fine-tuning occurs on gpus.
                 "device": "cuda",
@@ -47,7 +47,13 @@ def main() -> None:
                 # `lora_config` and `training_args` follow
                 # the standard lora and sfttrainer fields.
                 "fine_tune_config": {
-                    "lora_config": {"r": 16, "lora_alpha": 32},
+                    "lora_config": {
+                        "r": 16,
+                        "lora_alpha": 32,
+                        "lora_dropout": 0.05,
+                        "bias": "none",
+                        "task_type": "CAUSAL_LM",
+                    },
                     "training_args": {
                         "num_train_epochs": 5,
                         "logging_steps": 1,
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 6367382fe..e734aa34d 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -48,7 +48,6 @@ def __init__(
         output_dir: Optional[str] = None,
         device: Optional[torch.device] = None,
         local_model_path: Optional[str] = None,
-        local_tokenizer_path: Optional[str] = None,
         fine_tune_config: Optional[Dict[str, Any]] = None,
         **kwargs: Any,
     ) -> None:
@@ -72,7 +71,7 @@ def __init__(
             device (torch.device, optional): Device to run the model on.
                                              Default to GPU if available.
             local_model_path (str, optional): Local file path to a
-                                              pre-trained model.
+                                              pre-trained model and its tokenizer.
             fine_tune_config (dict, optional): Configuration for
                                                fine-tuning the model.
             **kwargs: Additional keyword arguments.
@@ -103,7 +102,7 @@ def __init__(
         )
         self.load_tokenizer(
             pretrained_model_name_or_path,
-            local_tokenizer_path=local_tokenizer_path,
+            local_model_path=local_model_path,
         )
 
         if data_path is not None:
@@ -250,6 +249,9 @@ def load_model(
         if bnb_config_default:
             bnb_config = BitsAndBytesConfig(**bnb_config_default)
 
+        self.lora_config = None
+        lora_config_default = {}
+
         try:
             if local_model_path is None:
                 self.model = AutoModelForCausalLM.from_pretrained(
@@ -263,6 +265,14 @@ def load_model(
                     ),
                     token=self.huggingface_token,
                 )
+
+                if fine_tune_config is not None:
+                    if fine_tune_config.get("lora_config") is not None:
+                        lora_config_default.update(fine_tune_config["lora_config"])
+                if lora_config_default != {}:
+                    self.lora_config = LoraConfig(**lora_config_default)
+                    self.model = get_peft_model(self.model, self.lora_config)
+
                 info_msg = (
                     f"Successfully loaded new model "
                     f"'{pretrained_model_name_or_path}' from "
@@ -280,6 +290,14 @@ def load_model(
                     ),
                     local_files_only=True,
                 )
+
+                if fine_tune_config is not None:
+                    if fine_tune_config.get("lora_config") is not None:
+                        lora_config_default.update(fine_tune_config["lora_config"])
+                if lora_config_default != {}:
+                    lora_config = LoraConfig(**lora_config_default)
+                    self.model = get_peft_model(self.model, lora_config)
+
                 info_msg = (
                     f"Successfully loaded new model "
                     f"'{pretrained_model_name_or_path}' from "
@@ -304,14 +322,15 @@ def load_model(
     def load_tokenizer(
         self,
         pretrained_model_name_or_path: Optional[str] = None,
-        local_tokenizer_path: Optional[str] = None,
+        local_model_path: Optional[str] = None,
     ) -> None:
         """
         Load the tokenizer from a local path.
 
         Arguments:
-            local_tokenizer_path (str): The file path to the
-                                        tokenizer to be loaded.
+            local_model_path (str): The file path to the
+                                    tokenizer to be loaded
+                                    (same as `local_model_path`).
             pretrained_model_name_or_path (str): An identifier
                                                 for the model on Huggingface.
             fine_tune_config (dict, optional): Configuration options for
@@ -325,7 +344,7 @@ def load_tokenizer(
         """
 
         try:
-            if local_tokenizer_path is None:
+            if local_model_path is None:
                 self.tokenizer = AutoTokenizer.from_pretrained(
                     pretrained_model_name_or_path,
                     token=self.huggingface_token,
@@ -338,13 +357,13 @@ def load_tokenizer(
 
             else:
                 self.tokenizer = AutoTokenizer.from_pretrained(
-                    local_tokenizer_path,
+                    local_model_path,
                 )
                 # log the successful tokenizer loading
                 logger.info(
                     f"Successfully loaded new tokenizer for model "
                     f"'{pretrained_model_name_or_path}'"
-                    f" from '{local_tokenizer_path}'",
+                    f" from '{local_model_path}'",
                 )
             self.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
 
@@ -354,7 +373,7 @@ def load_tokenizer(
             error_message = (
                 f"Failed to load tokenizer for model"
                 f" '{pretrained_model_name_or_path}' from "
-                f"'{local_tokenizer_path}': {e}"
+                f"'{local_model_path}': {e}"
             )
             logger.error(error_message)
 
@@ -484,9 +503,6 @@ def fine_tune_training(
 
         formatted_dataset = dataset_reduced.train_test_split(test_size=0.1)
 
-        lora_config = None
-        lora_config_default = {}
-
         training_defaults = {
             "per_device_train_batch_size": 1,
             "gradient_accumulation_steps": 4,
@@ -498,18 +514,12 @@ def fine_tune_training(
         }
 
         if fine_tune_config is not None:
-            if fine_tune_config.get("lora_config") is not None:
-                lora_config_default.update(fine_tune_config["lora_config"])
             if fine_tune_config.get("training_args") is not None:
                 training_defaults.update(fine_tune_config["training_args"])
 
         if output_dir is not None:
             training_defaults["output_dir"] = output_dir
 
-        if lora_config_default:
-            lora_config = LoraConfig(**lora_config_default)
-            model = get_peft_model(model, lora_config)
-
         collator = DataCollatorForCompletionOnlyLM(
             response_template=" ### Answer:",
             tokenizer=tokenizer,
@@ -524,7 +534,7 @@ def fine_tune_training(
             train_dataset=formatted_dataset["train"],
             eval_dataset=formatted_dataset["test"],
             **(
-                {"peft_config": lora_config} if lora_config is not None else {}
+                {"peft_config": self.lora_config} if self.lora_config is not None else {}
             ),
             args=trainer_args,
             max_seq_length=2048,
@@ -569,19 +579,4 @@ def fine_tune_training(
             model_path = os.path.join(os.path.dirname(__file__), model_name)
         trainer.save_model(model_path)
 
-        # save tokenizer
-        tokenizer_name_temp = model.config.name_or_path.split("/")[-1]
-        tokenizer_name = f"sft_{tokenizer_name_temp}_tokenizer_{time_string}"
-        if output_dir is not None:
-            tokenizer_path = os.path.join(
-                output_dir,
-                tokenizer_name,
-            )
-        else:
-            tokenizer_path = os.path.join(
-                os.path.dirname(__file__),
-                tokenizer_name,
-            )
-        tokenizer.save_pretrained(tokenizer_path)
-
-        return model
+        return model
\ No newline at end of file

From a27c3a82fe9bd508ac615fc9cc20ffcba59c27e2 Mon Sep 17 00:00:00 2001
From: Ze Yu Zhang <zyzhang1130@gmail.com>
Date: Fri, 7 Jun 2024 18:22:13 +0800
Subject: [PATCH 45/55] moved peft loading to ; removed independently saving
 tokenizer

---
 .../huggingface_model.py                      | 21 ++++++++++++-------
 1 file changed, 14 insertions(+), 7 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index e734aa34d..f2fb2cf30 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -71,7 +71,8 @@ def __init__(
             device (torch.device, optional): Device to run the model on.
                                              Default to GPU if available.
             local_model_path (str, optional): Local file path to a
-                                              pre-trained model and its tokenizer.
+                                              pre-trained model
+                                              and its tokenizer.
             fine_tune_config (dict, optional): Configuration for
                                                fine-tuning the model.
             **kwargs: Additional keyword arguments.
@@ -268,8 +269,10 @@ def load_model(
 
                 if fine_tune_config is not None:
                     if fine_tune_config.get("lora_config") is not None:
-                        lora_config_default.update(fine_tune_config["lora_config"])
-                if lora_config_default != {}:
+                        lora_config_default.update(
+                            fine_tune_config["lora_config"],
+                        )
+                if lora_config_default:
                     self.lora_config = LoraConfig(**lora_config_default)
                     self.model = get_peft_model(self.model, self.lora_config)
 
@@ -293,8 +296,10 @@ def load_model(
 
                 if fine_tune_config is not None:
                     if fine_tune_config.get("lora_config") is not None:
-                        lora_config_default.update(fine_tune_config["lora_config"])
-                if lora_config_default != {}:
+                        lora_config_default.update(
+                            fine_tune_config["lora_config"],
+                        )
+                if lora_config_default:
                     lora_config = LoraConfig(**lora_config_default)
                     self.model = get_peft_model(self.model, lora_config)
 
@@ -534,7 +539,9 @@ def fine_tune_training(
             train_dataset=formatted_dataset["train"],
             eval_dataset=formatted_dataset["test"],
             **(
-                {"peft_config": self.lora_config} if self.lora_config is not None else {}
+                {"peft_config": self.lora_config}
+                if self.lora_config is not None
+                else {}
             ),
             args=trainer_args,
             max_seq_length=2048,
@@ -579,4 +586,4 @@ def fine_tune_training(
             model_path = os.path.join(os.path.dirname(__file__), model_name)
         trainer.save_model(model_path)
 
-        return model
\ No newline at end of file
+        return model

From b5323461757edd7966d66326cbab33e7dbe1a7bd Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Mon, 24 Jun 2024 12:08:45 +0800
Subject: [PATCH 46/55] Update huggingface_model.py

updated according to the latest comments
---
 .../huggingface_model.py                        | 17 ++++++++++-------
 1 file changed, 10 insertions(+), 7 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index f2fb2cf30..1dbfcd846 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -79,6 +79,8 @@ def __init__(
         """
         super().__init__(config_name=config_name)
         self.model = None
+        self.lora_config = None
+        self.tokenizer = None
         self.max_length = max_length  # Set max_length as an attribute
         self.pretrained_model_name_or_path = pretrained_model_name_or_path
         script_path = os.path.abspath(__file__)
@@ -218,7 +220,7 @@ def format(
 
         return huggingface_msgs
 
-    def load_model(
+    def _load_model(
         self,
         pretrained_model_name_or_path: Optional[str] = None,
         local_model_path: Optional[str] = None,
@@ -250,7 +252,6 @@ def load_model(
         if bnb_config_default:
             bnb_config = BitsAndBytesConfig(**bnb_config_default)
 
-        self.lora_config = None
         lora_config_default = {}
 
         try:
@@ -324,7 +325,7 @@ def load_model(
 
             raise
 
-    def load_tokenizer(
+    def _load_tokenizer(
         self,
         pretrained_model_name_or_path: Optional[str] = None,
         local_model_path: Optional[str] = None,
@@ -384,7 +385,7 @@ def load_tokenizer(
 
             raise
 
-    def fine_tune(
+    def _fine_tune(
         self,
         data_path: Optional[str] = None,
         output_dir: Optional[str] = None,
@@ -426,7 +427,7 @@ def fine_tune(
             )
             raise
 
-    def formatting_prompts_func(
+    def _formatting_prompts_func(
         self,
         example: Dict[str, List[List[str]]],
     ) -> List[str]:
@@ -449,7 +450,7 @@ def formatting_prompts_func(
             output_texts.append(text)
         return output_texts
 
-    def fine_tune_training(
+    def _fine_tune_training(
         self,
         model: AutoModelForCausalLM,
         tokenizer: AutoTokenizer,
@@ -548,7 +549,9 @@ def fine_tune_training(
         )
 
         logger.info(
-            "fine-tuning model",
+            "Starting fine-tuning of the model '{model_name}' at {timestamp}",
+            model_name=self.model.config.name_or_path,
+            timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
         )
 
         trainer.train()

From d0bcc344ae508d4b7c88837dc77ba2b709a41ec4 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Mon, 24 Jun 2024 13:57:20 +0800
Subject: [PATCH 47/55] updated example
 `conversation_with_agent_with_finetuned_model` according to comments

---
 .../FinetuneDialogAgent.py                    | 26 ++++------
 ...rsation_with_agent_with_finetuned_model.py | 52 +++++++++++--------
 .../huggingface_model.py                      | 31 ++++++-----
 3 files changed, 59 insertions(+), 50 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
index b708ca81b..dd1ac2121 100644
--- a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
+++ b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
@@ -5,9 +5,7 @@
 capabilities with custom hyperparameters.
 """
 from typing import Any, Optional, Dict
-
 from loguru import logger
-
 from agentscope.agents import DialogAgent
 
 
@@ -45,7 +43,6 @@ def __init__(
         Note:
             Refer to `class DialogAgent(AgentBase)` for more information.
         """
-
         super().__init__(
             name,
             sys_prompt,
@@ -58,6 +55,7 @@ def load_model(
         self,
         pretrained_model_name_or_path: Optional[str] = None,
         local_model_path: Optional[str] = None,
+        fine_tune_config: Optional[Dict[str, Any]] = None,
     ) -> None:
         """
         Load a new model into the agent.
@@ -72,11 +70,11 @@ def load_model(
             Exception: If the model loading process fails or if the
                        model wrapper does not support dynamic loading.
         """
-
-        if hasattr(self.model, "load_model"):
-            self.model.load_model(
+        if hasattr(self.model, "_load_model"):
+            self.model._load_model(
                 pretrained_model_name_or_path,
                 local_model_path,
+                fine_tune_config
             )
         else:
             logger.error(
@@ -86,7 +84,7 @@ def load_model(
     def load_tokenizer(
         self,
         pretrained_model_name_or_path: Optional[str] = None,
-        local_tokenizer_path: Optional[str] = None,
+        local_model_path: Optional[str] = None,
     ) -> None:
         """
         Load a new tokenizer for the agent.
@@ -102,12 +100,8 @@ def load_tokenizer(
             Exception: If the model tokenizer process fails or if the
                        model wrapper does not support dynamic loading.
         """
-
-        if hasattr(self.model, "load_tokenizer"):
-            self.model.load_tokenizer(
-                pretrained_model_name_or_path,
-                local_tokenizer_path,
-            )
+        if hasattr(self.model, "_load_tokenizer"):
+            self.model._load_tokenizer(pretrained_model_name_or_path, local_model_path)
         else:
             logger.error("The model wrapper does not support dynamic loading.")
 
@@ -132,9 +126,7 @@ def fine_tune(
             Exception: If fine-tuning fails or if the
                        model wrapper does not support fine-tuning.
         """
-
-        if hasattr(self.model, "fine_tune"):
-            self.model.fine_tune(data_path, output_dir, fine_tune_config)
-            logger.info("Fine-tuning completed successfully.")
+        if hasattr(self.model, "_fine_tune"):
+            self.model._fine_tune(data_path, output_dir, fine_tune_config)
         else:
             logger.error("The model wrapper does not support fine-tuning.")
diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 79ba887a9..929a4aee9 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -85,31 +85,41 @@ def main() -> None:
         model_config_name="my_custom_model",
     )
 
-    # # (Optional) can load another model after
-    # # the agent has been instantiated if needed
-    # dialog_agent.load_model(
-    #     pretrained_model_name_or_path="google/gemma-7b",
-    #     local_model_path=None,
-    # )  # load model gemma-2b-it from Hugging Face
-    # dialog_agent.load_tokenizer(
-    #     pretrained_model_name_or_path="google/gemma-7b",
-    #     local_tokenizer_path=None,
-    # )  # load tokenizer for gemma-2b-it from Hugging Face
+    # (Optional) can load another model after
+    # the agent has been instantiated if needed
+    # (for `fine_tune_config` specify only
+    # `lora_config` and `bnb_config` if used)
+    dialog_agent.load_model(
+        pretrained_model_name_or_path="google/gemma-7b",
+        local_model_path=None,
+        fine_tune_config={
+            "lora_config": {"r": 24, "lora_alpha": 48},
+            "bnb_config": {
+                "load_in_4bit": True,
+                "bnb_4bit_use_double_quant": True,
+                "bnb_4bit_quant_type": "nf4",
+                "bnb_4bit_compute_dtype": "bfloat16",
+            },
+        },
+    )  # load model from Hugging Face
 
-    # fine-tune loaded model with lima dataset
-    # with default hyperparameters
-    # dialog_agent.fine_tune(data_path="GAIR/lima")
+    dialog_agent.load_tokenizer(
+        pretrained_model_name_or_path="google/gemma-7b",
+        local_model_path=None,
+    )  # load tokenizer
 
     # fine-tune loaded model with lima dataset
     # with customized hyperparameters
-    # (`fine_tune_config` argument is optional. Defaults to None.)
-    # dialog_agent.fine_tune(
-    #     "GAIR/lima",
-    #     fine_tune_config={
-    #         "lora_config": {"r": 24, "lora_alpha": 48},
-    #         "training_args": {"max_steps": 300, "logging_steps": 3},
-    #     },
-    # )
+    # (`fine_tune_config` argument is optional
+    # (specify only `lora_config` and
+    # `training_args` if used). Defaults to None.)
+    dialog_agent.fine_tune(
+        "GAIR/lima",
+        fine_tune_config={
+            "lora_config": {"r": 24, "lora_alpha": 48},
+            "training_args": {"max_steps": 300, "logging_steps": 3},
+        },
+    )
 
     user_agent = UserAgent()
 
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 1dbfcd846..4f892090e 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -79,8 +79,6 @@ def __init__(
         """
         super().__init__(config_name=config_name)
         self.model = None
-        self.lora_config = None
-        self.tokenizer = None
         self.max_length = max_length  # Set max_length as an attribute
         self.pretrained_model_name_or_path = pretrained_model_name_or_path
         script_path = os.path.abspath(__file__)
@@ -98,18 +96,18 @@ def __init__(
             self.device_map = device
             self.device = device
 
-        self.load_model(
+        self._load_model(
             pretrained_model_name_or_path,
             local_model_path=local_model_path,
             fine_tune_config=fine_tune_config,
         )
-        self.load_tokenizer(
+        self._load_tokenizer(
             pretrained_model_name_or_path,
             local_model_path=local_model_path,
         )
 
         if data_path is not None:
-            self.model = self.fine_tune_training(
+            self.model = self._fine_tune_training(
                 self.model,
                 self.tokenizer,
                 data_path,
@@ -252,6 +250,7 @@ def _load_model(
         if bnb_config_default:
             bnb_config = BitsAndBytesConfig(**bnb_config_default)
 
+        self.lora_config = None
         lora_config_default = {}
 
         try:
@@ -409,7 +408,7 @@ def _fine_tune(
             or internal errors during the training process.
         """
         try:
-            self.model = self.fine_tune_training(
+            self.model = self._fine_tune_training(
                 self.model,
                 self.tokenizer,
                 data_path,
@@ -519,9 +518,20 @@ def _fine_tune_training(
             "logging_steps": 1,
         }
 
+        self.lora_config = None
+        lora_config_default = {}
+
         if fine_tune_config is not None:
             if fine_tune_config.get("training_args") is not None:
                 training_defaults.update(fine_tune_config["training_args"])
+            if fine_tune_config.get("lora_config") is not None:
+                lora_config_default.update(
+                    fine_tune_config["lora_config"],
+                )
+
+        if lora_config_default:
+            self.lora_config = LoraConfig(**lora_config_default)
+            self.model = get_peft_model(self.model, self.lora_config)       
 
         if output_dir is not None:
             training_defaults["output_dir"] = output_dir
@@ -535,7 +545,7 @@ def _fine_tune_training(
 
         trainer = SFTTrainer(
             model,
-            formatting_func=self.formatting_prompts_func,
+            formatting_func=self._formatting_prompts_func,
             data_collator=collator,
             train_dataset=formatted_dataset["train"],
             eval_dataset=formatted_dataset["test"],
@@ -548,11 +558,8 @@ def _fine_tune_training(
             max_seq_length=2048,
         )
 
-        logger.info(
-            "Starting fine-tuning of the model '{model_name}' at {timestamp}",
-            model_name=self.model.config.name_or_path,
-            timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
-        )
+        logger.info("Starting fine-tuning of the model '{model_name}' at {timestamp}", model_name=self.model.config.name_or_path, timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
+
 
         trainer.train()
 

From 21a9f21683587c3cf682574d460c512af4f06e3d Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Mon, 24 Jun 2024 14:08:00 +0800
Subject: [PATCH 48/55] bug fixing

---
 .../FinetuneDialogAgent.py                           | 10 +++++-----
 .../huggingface_model.py                             | 12 ++++++------
 2 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
index dd1ac2121..a8b919a1f 100644
--- a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
+++ b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
@@ -43,7 +43,7 @@ def __init__(
         Note:
             Refer to `class DialogAgent(AgentBase)` for more information.
         """
-        super().__init__(
+        super().__init__( # pylint: disable=useless-parent-delegation
             name,
             sys_prompt,
             model_config_name,
@@ -100,8 +100,8 @@ def load_tokenizer(
             Exception: If the model tokenizer process fails or if the
                        model wrapper does not support dynamic loading.
         """
-        if hasattr(self.model, "_load_tokenizer"):
-            self.model._load_tokenizer(pretrained_model_name_or_path, local_model_path)
+        if hasattr(self.model, "load_tokenizer"):
+            self.model.load_tokenizer(pretrained_model_name_or_path, local_model_path)
         else:
             logger.error("The model wrapper does not support dynamic loading.")
 
@@ -126,7 +126,7 @@ def fine_tune(
             Exception: If fine-tuning fails or if the
                        model wrapper does not support fine-tuning.
         """
-        if hasattr(self.model, "_fine_tune"):
-            self.model._fine_tune(data_path, output_dir, fine_tune_config)
+        if hasattr(self.model, "fine_tune"):
+            self.model.fine_tune(data_path, output_dir, fine_tune_config)
         else:
             logger.error("The model wrapper does not support fine-tuning.")
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 4f892090e..2cac69aaa 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -101,13 +101,13 @@ def __init__(
             local_model_path=local_model_path,
             fine_tune_config=fine_tune_config,
         )
-        self._load_tokenizer(
+        self.load_tokenizer(
             pretrained_model_name_or_path,
             local_model_path=local_model_path,
         )
 
         if data_path is not None:
-            self.model = self._fine_tune_training(
+            self.model = self.fine_tune_training(
                 self.model,
                 self.tokenizer,
                 data_path,
@@ -324,7 +324,7 @@ def _load_model(
 
             raise
 
-    def _load_tokenizer(
+    def load_tokenizer(
         self,
         pretrained_model_name_or_path: Optional[str] = None,
         local_model_path: Optional[str] = None,
@@ -384,7 +384,7 @@ def _load_tokenizer(
 
             raise
 
-    def _fine_tune(
+    def fine_tune(
         self,
         data_path: Optional[str] = None,
         output_dir: Optional[str] = None,
@@ -408,7 +408,7 @@ def _fine_tune(
             or internal errors during the training process.
         """
         try:
-            self.model = self._fine_tune_training(
+            self.model = self.fine_tune_training(
                 self.model,
                 self.tokenizer,
                 data_path,
@@ -449,7 +449,7 @@ def _formatting_prompts_func(
             output_texts.append(text)
         return output_texts
 
-    def _fine_tune_training(
+    def fine_tune_training(
         self,
         model: AutoModelForCausalLM,
         tokenizer: AutoTokenizer,

From 97e633010b5bb66eea00e277ed835b85f0a63098 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Mon, 24 Jun 2024 14:15:02 +0800
Subject: [PATCH 49/55] bug fixing

---
 .../FinetuneDialogAgent.py                                 | 7 ++++---
 .../huggingface_model.py                                   | 4 ++--
 tests/model_test.py                                        | 2 +-
 3 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
index a8b919a1f..68ff08b2e 100644
--- a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
+++ b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
@@ -43,7 +43,8 @@ def __init__(
         Note:
             Refer to `class DialogAgent(AgentBase)` for more information.
         """
-        super().__init__( # pylint: disable=useless-parent-delegation
+        # pylint: disable=useless-parent-delegation
+        super().__init__(
             name,
             sys_prompt,
             model_config_name,
@@ -70,8 +71,8 @@ def load_model(
             Exception: If the model loading process fails or if the
                        model wrapper does not support dynamic loading.
         """
-        if hasattr(self.model, "_load_model"):
-            self.model._load_model(
+        if hasattr(self.model, "load_model"):
+            self.model.load_model(
                 pretrained_model_name_or_path,
                 local_model_path,
                 fine_tune_config
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 2cac69aaa..49ea6d522 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -96,7 +96,7 @@ def __init__(
             self.device_map = device
             self.device = device
 
-        self._load_model(
+        self.load_model(
             pretrained_model_name_or_path,
             local_model_path=local_model_path,
             fine_tune_config=fine_tune_config,
@@ -218,7 +218,7 @@ def format(
 
         return huggingface_msgs
 
-    def _load_model(
+    def load_model(
         self,
         pretrained_model_name_or_path: Optional[str] = None,
         local_model_path: Optional[str] = None,
diff --git a/tests/model_test.py b/tests/model_test.py
index f361b7f0d..e3d22f41d 100644
--- a/tests/model_test.py
+++ b/tests/model_test.py
@@ -55,7 +55,7 @@ def test_model_registry(self) -> None:
         )
 
     @patch("loguru.logger.warning")
-    def test_load_model_configs(self, mock_logging: MagicMock) -> None:
+    def testload_model_configs(self, mock_logging: MagicMock) -> None:
         """Test to load model configs"""
         configs = [
             {

From b4a3849308e4f312a25e08a004e93cbc84814fb7 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Mon, 24 Jun 2024 14:18:52 +0800
Subject: [PATCH 50/55] bug fixing

---
 .../FinetuneDialogAgent.py                               | 7 +++++--
 .../huggingface_model.py                                 | 9 ++++++---
 2 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
index 68ff08b2e..f3167fde4 100644
--- a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
+++ b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
@@ -75,7 +75,7 @@ def load_model(
             self.model.load_model(
                 pretrained_model_name_or_path,
                 local_model_path,
-                fine_tune_config
+                fine_tune_config,
             )
         else:
             logger.error(
@@ -102,7 +102,10 @@ def load_tokenizer(
                        model wrapper does not support dynamic loading.
         """
         if hasattr(self.model, "load_tokenizer"):
-            self.model.load_tokenizer(pretrained_model_name_or_path, local_model_path)
+            self.model.load_tokenizer(
+                pretrained_model_name_or_path,
+                local_model_path,
+            )
         else:
             logger.error("The model wrapper does not support dynamic loading.")
 
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 49ea6d522..4dbab7a52 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -531,7 +531,7 @@ def fine_tune_training(
 
         if lora_config_default:
             self.lora_config = LoraConfig(**lora_config_default)
-            self.model = get_peft_model(self.model, self.lora_config)       
+            self.model = get_peft_model(self.model, self.lora_config)
 
         if output_dir is not None:
             training_defaults["output_dir"] = output_dir
@@ -558,8 +558,11 @@ def fine_tune_training(
             max_seq_length=2048,
         )
 
-        logger.info("Starting fine-tuning of the model '{model_name}' at {timestamp}", model_name=self.model.config.name_or_path, timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
-
+        logger.info(
+            "Starting fine-tuning of the model '{model_name}' at {timestamp}",
+            model_name=self.model.config.name_or_path,
+            timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+        )
 
         trainer.train()
 

From 799cac5574859795014eef0047f0ff63f8935d6e Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 17 Jul 2024 19:40:35 +0800
Subject: [PATCH 51/55] changed to using `chat_template` format for finetuning.
 Resolve the bug for continual finetuning.

---
 ...rsation_with_agent_with_finetuned_model.py |  62 +++--
 .../huggingface_model.py                      | 223 ++++++++++++------
 2 files changed, 184 insertions(+), 101 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 929a4aee9..2a1c34d80 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -9,6 +9,7 @@
 Features include model and tokenizer loading,
 and fine-tuning on the lima dataset with adjustable parameters.
 """
+
 # pylint: disable=unused-import
 from huggingface_model import HuggingFaceWrapper
 from FinetuneDialogAgent import FinetuneDialogAgent
@@ -30,9 +31,8 @@ def main() -> None:
                 # Or another generative model of your choice.
                 # Needed from loading from Hugging Face.
                 "pretrained_model_name_or_path": "google/gemma-7b",
-                # "local_model_path": , # Specify your local model path
-                # "local_tokenizer_path":,  # Specify your local tokenizer path
-                "max_length": 128,
+                # "local_model_path": "", # Specify your local model path
+                "max_length": 256,
                 # Device for inference. Fine-tuning occurs on gpus.
                 "device": "cuda",
                 # Specify a Hugging Face data path if you
@@ -46,8 +46,13 @@ def main() -> None:
                 # fine-tuning method. Defaults to None.
                 # `lora_config` and `training_args` follow
                 # the standard lora and sfttrainer fields.
+                # "lora_config" shouldn't be specified if 
+                # loading a model saved as lora model
+                # '"continue_lora_finetuning": True' if
+                # loading a model saved as lora model
                 "fine_tune_config": {
-                    "lora_config": {
+                    "continue_lora_finetuning": False,                                                        
+                    "lora_config": {               
                         "r": 16,
                         "lora_alpha": 32,
                         "lora_dropout": 0.05,
@@ -56,14 +61,16 @@ def main() -> None:
                     },
                     "training_args": {
                         "num_train_epochs": 5,
+                        # "max_steps": 100,
                         "logging_steps": 1,
+                        # "learning_rate": 5e-07
                     },
-                    "bnb_config": {
-                        "load_in_4bit": True,
-                        "bnb_4bit_use_double_quant": True,
-                        "bnb_4bit_quant_type": "nf4",
-                        "bnb_4bit_compute_dtype": "bfloat16",
-                    },
+                    # "bnb_config": {
+                    #     "load_in_8bit": True,
+                        # "bnb_4bit_use_double_quant": True,
+                        # "bnb_4bit_quant_type": "nf4",
+                        # "bnb_4bit_compute_dtype": "bfloat16",
+                    # },
                 },
             },
         ],
@@ -78,8 +85,7 @@ def main() -> None:
     dialog_agent = FinetuneDialogAgent(
         name="Assistant",
         sys_prompt=(
-            "Explain in simple terms how the attention mechanism of "
-            "a transformer model works."
+            "You're a helpful assistant."
         ),
         # Use your custom model config name here
         model_config_name="my_custom_model",
@@ -91,32 +97,36 @@ def main() -> None:
     # `lora_config` and `bnb_config` if used)
     dialog_agent.load_model(
         pretrained_model_name_or_path="google/gemma-7b",
-        local_model_path=None,
+        # local_model_path="",
         fine_tune_config={
-            "lora_config": {"r": 24, "lora_alpha": 48},
-            "bnb_config": {
-                "load_in_4bit": True,
-                "bnb_4bit_use_double_quant": True,
-                "bnb_4bit_quant_type": "nf4",
-                "bnb_4bit_compute_dtype": "bfloat16",
-            },
+            # "bnb_config": {
+            #     "load_in_4bit": True,
+            #     "bnb_4bit_use_double_quant": True,
+            #     "bnb_4bit_quant_type": "nf4",
+            #     "bnb_4bit_compute_dtype": "bfloat16",
+            # },
         },
     )  # load model from Hugging Face
 
     dialog_agent.load_tokenizer(
         pretrained_model_name_or_path="google/gemma-7b",
-        local_model_path=None,
+        # local_model_path="",
     )  # load tokenizer
 
     # fine-tune loaded model with lima dataset
     # with customized hyperparameters
-    # (`fine_tune_config` argument is optional
-    # (specify only `lora_config` and
-    # `training_args` if used). Defaults to None.)
+    # `fine_tune_config` argument is optional
+    # specify only `lora_config` and
+    # `training_args` if used). Defaults to None.
+    # "lora_config" shouldn't be specified if 
+    # loading a model saved as lora model
+    # '"continue_lora_finetuning": True' if
+    # loading a model saved as lora model
     dialog_agent.fine_tune(
         "GAIR/lima",
         fine_tune_config={
-            "lora_config": {"r": 24, "lora_alpha": 48},
+            "continue_lora_finetuning": True,
+            # "lora_config": {"r": 24, "lora_alpha": 48}, 
             "training_args": {"max_steps": 300, "logging_steps": 3},
         },
     )
@@ -126,7 +136,7 @@ def main() -> None:
     # Start the conversation between user and assistant
     x = None
     while x is None or x.content != "exit":
-        x = sequentialpipeline([dialog_agent, user_agent], x)
+        x = sequentialpipeline([user_agent, dialog_agent], x)
 
 
 if __name__ == "__main__":
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index 4dbab7a52..a844b22ec 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -6,22 +6,26 @@
 Key features include handling model and tokenizer operations,
 adapting to specialized datasets, and robust error management.
 """
-from typing import Sequence, Any, Union, List, Optional, Dict
+from typing import Sequence, Any, Union, List, Optional, Dict, Literal, Tuple
 import os
 from datetime import datetime
 import json
+from dataclasses import dataclass
 
 import torch
 from transformers import (
     AutoModelForCausalLM,
     AutoTokenizer,
     BitsAndBytesConfig,
+    PreTrainedModel,
+    PreTrainedTokenizer,
 )
 import transformers
 from peft import LoraConfig
 from peft import get_peft_model
+from peft import PeftModel, PeftConfig
 from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
-from datasets import load_dataset
+from datasets import load_dataset, Dataset
 from loguru import logger
 from dotenv import load_dotenv
 
@@ -30,6 +34,7 @@
 from agentscope.utils.tools import _convert_to_str
 
 
+
 class HuggingFaceWrapper(ModelWrapperBase):
     """Wrapper for a Hugging Face transformer model.
 
@@ -81,6 +86,7 @@ def __init__(
         self.model = None
         self.max_length = max_length  # Set max_length as an attribute
         self.pretrained_model_name_or_path = pretrained_model_name_or_path
+        self.local_model_path = local_model_path
         script_path = os.path.abspath(__file__)
         script_dir = os.path.dirname(script_path)
         dotenv_path = os.path.join(script_dir, ".env")
@@ -105,7 +111,7 @@ def __init__(
             pretrained_model_name_or_path,
             local_model_path=local_model_path,
         )
-
+        
         if data_path is not None:
             self.model = self.fine_tune_training(
                 self.model,
@@ -142,15 +148,24 @@ def __call__(
             Exception: If an error occurs during text generation.
         """
 
+        if self.local_model_path is not None:
+            if self.tokenizer.chat_template is not None:
+                if self.model.get_input_embeddings().weight.size()[0] != len(self.tokenizer):
+                    self.model, self.tokenizer = self.setup_chat_format(self.model, self.tokenizer)
+                    loaded_embed_tokens_weights = torch.load(self.local_model_path+'/embed_tokens_weights.pt')
+                    self.model.get_input_embeddings().weight.data.copy_(loaded_embed_tokens_weights)
+                    del loaded_embed_tokens_weights
+                    torch.cuda.empty_cache()
+            else:
+                if self.tokenizer.chat_template is None:
+                    self.model, self.tokenizer = self.setup_chat_format(self.model, self.tokenizer)
+        else:
+            if self.tokenizer.chat_template is None:
+                self.model, self.tokenizer = self.setup_chat_format(self.model, self.tokenizer)
+
         try:
-            # Tokenize the input text
-            concatenated_input = "\n ".join(
-                [f"{d.get('role')}: {d['content']}" for d in inputs],
-            )
-            input_ids = self.tokenizer.encode(
-                f"{concatenated_input}\n assistent: ",
-                return_tensors="pt",
-            )
+
+            input_ids = self.tokenizer.apply_chat_template(inputs, tokenize=True, add_generation_prompt=True, return_tensors="pt")
             # Generate response using the model
             outputs = self.model.generate(
                 input_ids.to(self.device),
@@ -240,7 +255,7 @@ def load_model(
                        incorrect model ID,
                        or network issues while fetching the model.
         """
-
+        self.local_model_path = local_model_path
         bnb_config = None
         bnb_config_default = {}
 
@@ -250,9 +265,6 @@ def load_model(
         if bnb_config_default:
             bnb_config = BitsAndBytesConfig(**bnb_config_default)
 
-        self.lora_config = None
-        lora_config_default = {}
-
         try:
             if local_model_path is None:
                 self.model = AutoModelForCausalLM.from_pretrained(
@@ -267,15 +279,6 @@ def load_model(
                     token=self.huggingface_token,
                 )
 
-                if fine_tune_config is not None:
-                    if fine_tune_config.get("lora_config") is not None:
-                        lora_config_default.update(
-                            fine_tune_config["lora_config"],
-                        )
-                if lora_config_default:
-                    self.lora_config = LoraConfig(**lora_config_default)
-                    self.model = get_peft_model(self.model, self.lora_config)
-
                 info_msg = (
                     f"Successfully loaded new model "
                     f"'{pretrained_model_name_or_path}' from "
@@ -294,15 +297,6 @@ def load_model(
                     local_files_only=True,
                 )
 
-                if fine_tune_config is not None:
-                    if fine_tune_config.get("lora_config") is not None:
-                        lora_config_default.update(
-                            fine_tune_config["lora_config"],
-                        )
-                if lora_config_default:
-                    lora_config = LoraConfig(**lora_config_default)
-                    self.model = get_peft_model(self.model, lora_config)
-
                 info_msg = (
                     f"Successfully loaded new model "
                     f"'{pretrained_model_name_or_path}' from "
@@ -338,16 +332,13 @@ def load_tokenizer(
                                     (same as `local_model_path`).
             pretrained_model_name_or_path (str): An identifier
                                                 for the model on Huggingface.
-            fine_tune_config (dict, optional): Configuration options for
-                                               fine-tuning the model,
-                                               including QLoRA and training
-                                               arguments.
+
         Raises:
             Exception: If the tokenizer cannot be loaded from the
             given path or identifier. Possible reasons include file not found,
             incorrect model ID, or network issues while fetching the tokenizer.
         """
-
+        self.local_model_path = local_model_path
         try:
             if local_model_path is None:
                 self.tokenizer = AutoTokenizer.from_pretrained(
@@ -370,7 +361,7 @@ def load_tokenizer(
                     f"'{pretrained_model_name_or_path}'"
                     f" from '{local_model_path}'",
                 )
-            self.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
+            self.tokenizer.padding_side = 'right'
 
         except Exception as e:
             # Handle exceptions during model loading,
@@ -383,6 +374,60 @@ def load_tokenizer(
             logger.error(error_message)
 
             raise
+    
+    def setup_chat_format(
+        self,
+        model: PreTrainedModel,
+        tokenizer: PreTrainedTokenizer,
+        resize_to_multiple_of: Optional[int] = None,
+    ) -> Tuple[PreTrainedModel, PreTrainedTokenizer]:
+        """
+        Setup chat format by adding special tokens to the tokenizer, setting the correct format, and extending the embedding layer of the model based on the new special tokens.
+
+        Args:
+        model (`~transformers.PreTrainedModel`): The model to be modified.
+        tokenizer (`~transformers.PreTrainedTokenizer`): The tokenizer to be modified.
+        format (`Optional[Literal["chatml"]]`): The format to be set. Defaults to "chatml".
+        resize_to_multiple_of (`Optional[int]`): Number to resize the embedding layer to. Defaults to None.
+        Returns:
+        model (`~transformers.PreTrainedModel`): The modified model.
+        tokenizer (`~transformers.PreTrainedTokenizer`): The modified tokenizer.
+        """
+        
+
+        # set special tokens and them
+        tokenizer.add_special_tokens({"additional_special_tokens": ['<|system|>', '<|user|>', '<|assistant|>']})
+        # set chat format for tokenizer
+        tokenizer.chat_template = """{% for message in messages %}
+        {% if message['role'] == 'user' %}
+        {{ '<|user|>\n' + message['content'] + eos_token }}
+        {% elif message['role'] == 'system' %}
+        {{ '<|system|>\n' + message['content'] + eos_token }}
+        {% elif message['role'] == 'assistant' %}
+        {{ '<|assistant|>\n'  + message['content'] + eos_token }}
+        {% endif %}
+        {% if loop.last and add_generation_prompt %}
+        {{ '<|assistant|>' }}
+        {% endif %}
+        {% endfor %}"""
+
+        # resize embedding layer to a multiple of 64, https://x.com/karpathy/status/1621578354024677377
+        model.resize_token_embeddings(
+            len(tokenizer), pad_to_multiple_of=resize_to_multiple_of if resize_to_multiple_of is not None else None
+        )
+        # Update the model config to use the new eos & bos tokens
+        if getattr(model, "config", None) is not None:
+            model.config.pad_token_id = tokenizer.pad_token_id
+            model.config.bos_token_id = tokenizer.bos_token_id
+            model.config.eos_token_id = tokenizer.eos_token_id
+        # Update the generation config to use the new eos & bos token
+        if getattr(model, "generation_config", None) is not None:
+            model.generation_config.bos_token_id = tokenizer.bos_token_id
+            model.generation_config.eos_token_id = tokenizer.eos_token_id
+            model.generation_config.pad_token_id = tokenizer.pad_token_id
+
+        return model, tokenizer
+
 
     def fine_tune(
         self,
@@ -498,21 +543,46 @@ def fine_tune_training(
             as part of the log/model fodler name.
         """
 
+        if self.local_model_path is not None:
+            if tokenizer.chat_template is not None:
+                if model.get_input_embeddings().weight.size()[0] != len(tokenizer):
+                    model, tokenizer = self.setup_chat_format(model, tokenizer)
+                    loaded_embed_tokens_weights = torch.load(self.local_model_path+'/embed_tokens_weights.pt')
+                    model.get_input_embeddings().weight.data.copy_(loaded_embed_tokens_weights)
+                    del loaded_embed_tokens_weights
+                    torch.cuda.empty_cache()
+            else:
+                if tokenizer.chat_template is None:
+                    model, tokenizer = self.setup_chat_format(model, tokenizer)
+        else:
+            if tokenizer.chat_template is None:
+                model, tokenizer = self.setup_chat_format(model, tokenizer)
+
         dataset = load_dataset(data_path, split="train", token=token)
 
-        # filter out input sequences that are longer than certain threshold
+        # # filter out input sequences that are longer than certain threshold
         dataset_reduced = dataset.filter(
             lambda x: len(x["conversations"][0] + x["conversations"][1])
-            <= 1000,
+            <=3000,
         )
 
-        formatted_dataset = dataset_reduced.train_test_split(test_size=0.1)
+        # Function to reformat a single row
+        def reformat_row(row):
+            return [
+                    {"role": "system", "content": "You're a helpful assistant."},
+                    {"role": "user", "content": row['conversations'][0]},
+                    {"role": "assistant", "content": row['conversations'][1]}
+            ]
+
+        # Apply the reformatting function to each row in the dataset
+        formatted_dataset = [reformat_row(row) for row in dataset_reduced]
+        formatted_dataset = Dataset.from_dict({"messages": formatted_dataset})
 
         training_defaults = {
-            "per_device_train_batch_size": 1,
-            "gradient_accumulation_steps": 4,
+            "per_device_train_batch_size": 4,
+            "gradient_accumulation_steps": 1,
             "gradient_checkpointing": False,
-            "num_train_epochs": 5,
+            "num_train_epochs": 3,
             "output_dir": "./",
             "optim": "paged_adamw_8bit",
             "logging_steps": 1,
@@ -529,9 +599,17 @@ def fine_tune_training(
                     fine_tune_config["lora_config"],
                 )
 
-        if lora_config_default:
-            self.lora_config = LoraConfig(**lora_config_default)
-            self.model = get_peft_model(self.model, self.lora_config)
+        if fine_tune_config.get("continue_lora_finetuning") is True:
+                self.lora_config = PeftConfig.from_pretrained(self.local_model_path)
+                model = PeftModel.from_pretrained(model, self.local_model_path)
+                # unfreeze lora parameters. Assuming 'lora' is in the layer name.
+                for name, param in model.named_parameters(): 
+                    if "lora" in name:
+                        param.requires_grad = True
+        else:
+            if lora_config_default:
+                self.lora_config = LoraConfig(**lora_config_default)
+                model = get_peft_model(model, self.lora_config)
 
         if output_dir is not None:
             training_defaults["output_dir"] = output_dir
@@ -543,51 +621,38 @@ def fine_tune_training(
 
         trainer_args = transformers.TrainingArguments(**training_defaults)
 
+        
         trainer = SFTTrainer(
-            model,
-            formatting_func=self._formatting_prompts_func,
-            data_collator=collator,
-            train_dataset=formatted_dataset["train"],
-            eval_dataset=formatted_dataset["test"],
+            model=model,
+            tokenizer=tokenizer,
+            train_dataset=formatted_dataset,
+            eval_dataset=formatted_dataset,
             **(
                 {"peft_config": self.lora_config}
                 if self.lora_config is not None
                 else {}
             ),
             args=trainer_args,
-            max_seq_length=2048,
+            max_seq_length=4096,
         )
 
         logger.info(
             "Starting fine-tuning of the model '{model_name}' at {timestamp}",
-            model_name=self.model.config.name_or_path,
+            model_name=model.config.name_or_path,
             timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
         )
 
-        trainer.train()
+        try:
+            trainer.train()
+        except Exception as e:
+            import traceback
+            logger.error(f"Error during training: {e}")
+            traceback.print_exc()
+            raise
 
         now = datetime.now()
         time_string = now.strftime("%Y-%m-%d_%H-%M-%S")
 
-        if output_dir is not None:
-            os.makedirs(output_dir, exist_ok=True)
-
-        # Specify the filename
-        log_name_temp = model.config.name_or_path.split("/")[-1]
-        log_name = f"{log_name_temp}_{time_string}_log_history.json"
-        log_path = os.path.join(os.path.dirname(__file__), log_name)
-
-        # log training history
-        if output_dir is not None:
-            with open(
-                os.path.join(output_dir, log_name),
-                "w",
-                encoding="utf-8",
-            ) as f:
-                json.dump(trainer.state.log_history, f)
-        else:
-            with open(log_path, "w", encoding="utf-8") as f:
-                json.dump(trainer.state.log_history, f)
 
         # save model
         model_name = (
@@ -599,4 +664,12 @@ def fine_tune_training(
             model_path = os.path.join(os.path.dirname(__file__), model_name)
         trainer.save_model(model_path)
 
+        # save token embeddings because it is resized due to the addition of new special tokens
+        embed_tokens_weights = model.get_input_embeddings().weight.data.clone().detach()
+        torch.save(embed_tokens_weights, model_path+'/embed_tokens_weights.pt')
+
+        with open(os.path.join(model_path, "log_history.json"), "w", encoding="utf-8") as f:
+            json.dump(trainer.state.log_history, f)
+
         return model
+

From a13b219962154103b7f55603a86d0921fc896b11 Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 17 Jul 2024 20:27:07 +0800
Subject: [PATCH 52/55] reformatted according to pre-commit

---
 ...rsation_with_agent_with_finetuned_model.py |  20 +-
 .../huggingface_model.py                      | 233 ++++++++++++------
 2 files changed, 167 insertions(+), 86 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 2a1c34d80..7e297aee9 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -46,13 +46,13 @@ def main() -> None:
                 # fine-tuning method. Defaults to None.
                 # `lora_config` and `training_args` follow
                 # the standard lora and sfttrainer fields.
-                # "lora_config" shouldn't be specified if 
+                # "lora_config" shouldn't be specified if
                 # loading a model saved as lora model
                 # '"continue_lora_finetuning": True' if
                 # loading a model saved as lora model
                 "fine_tune_config": {
-                    "continue_lora_finetuning": False,                                                        
-                    "lora_config": {               
+                    "continue_lora_finetuning": False,
+                    "lora_config": {
                         "r": 16,
                         "lora_alpha": 32,
                         "lora_dropout": 0.05,
@@ -67,9 +67,9 @@ def main() -> None:
                     },
                     # "bnb_config": {
                     #     "load_in_8bit": True,
-                        # "bnb_4bit_use_double_quant": True,
-                        # "bnb_4bit_quant_type": "nf4",
-                        # "bnb_4bit_compute_dtype": "bfloat16",
+                    # "bnb_4bit_use_double_quant": True,
+                    # "bnb_4bit_quant_type": "nf4",
+                    # "bnb_4bit_compute_dtype": "bfloat16",
                     # },
                 },
             },
@@ -84,9 +84,7 @@ def main() -> None:
     # Init agents with the custom model
     dialog_agent = FinetuneDialogAgent(
         name="Assistant",
-        sys_prompt=(
-            "You're a helpful assistant."
-        ),
+        sys_prompt=("You're a helpful assistant."),
         # Use your custom model config name here
         model_config_name="my_custom_model",
     )
@@ -118,7 +116,7 @@ def main() -> None:
     # `fine_tune_config` argument is optional
     # specify only `lora_config` and
     # `training_args` if used). Defaults to None.
-    # "lora_config" shouldn't be specified if 
+    # "lora_config" shouldn't be specified if
     # loading a model saved as lora model
     # '"continue_lora_finetuning": True' if
     # loading a model saved as lora model
@@ -126,7 +124,7 @@ def main() -> None:
         "GAIR/lima",
         fine_tune_config={
             "continue_lora_finetuning": True,
-            # "lora_config": {"r": 24, "lora_alpha": 48}, 
+            # "lora_config": {"r": 24, "lora_alpha": 48},
             "training_args": {"max_steps": 300, "logging_steps": 3},
         },
     )
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index a844b22ec..c3b0e0b01 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -6,11 +6,10 @@
 Key features include handling model and tokenizer operations,
 adapting to specialized datasets, and robust error management.
 """
-from typing import Sequence, Any, Union, List, Optional, Dict, Literal, Tuple
+from typing import Sequence, Any, Union, List, Optional, Dict, Tuple
 import os
 from datetime import datetime
 import json
-from dataclasses import dataclass
 
 import torch
 from transformers import (
@@ -24,7 +23,7 @@
 from peft import LoraConfig
 from peft import get_peft_model
 from peft import PeftModel, PeftConfig
-from trl import SFTTrainer, DataCollatorForCompletionOnlyLM
+from trl import SFTTrainer
 from datasets import load_dataset, Dataset
 from loguru import logger
 from dotenv import load_dotenv
@@ -34,7 +33,6 @@
 from agentscope.utils.tools import _convert_to_str
 
 
-
 class HuggingFaceWrapper(ModelWrapperBase):
     """Wrapper for a Hugging Face transformer model.
 
@@ -84,6 +82,7 @@ def __init__(
         """
         super().__init__(config_name=config_name)
         self.model = None
+        self.tokenizer = None
         self.max_length = max_length  # Set max_length as an attribute
         self.pretrained_model_name_or_path = pretrained_model_name_or_path
         self.local_model_path = local_model_path
@@ -111,7 +110,7 @@ def __init__(
             pretrained_model_name_or_path,
             local_model_path=local_model_path,
         )
-        
+
         if data_path is not None:
             self.model = self.fine_tune_training(
                 self.model,
@@ -147,25 +146,21 @@ def __call__(
         Raises:
             Exception: If an error occurs during text generation.
         """
-
-        if self.local_model_path is not None:
-            if self.tokenizer.chat_template is not None:
-                if self.model.get_input_embeddings().weight.size()[0] != len(self.tokenizer):
-                    self.model, self.tokenizer = self.setup_chat_format(self.model, self.tokenizer)
-                    loaded_embed_tokens_weights = torch.load(self.local_model_path+'/embed_tokens_weights.pt')
-                    self.model.get_input_embeddings().weight.data.copy_(loaded_embed_tokens_weights)
-                    del loaded_embed_tokens_weights
-                    torch.cuda.empty_cache()
-            else:
-                if self.tokenizer.chat_template is None:
-                    self.model, self.tokenizer = self.setup_chat_format(self.model, self.tokenizer)
+        if self.tokenizer is not None:
+            self.model, self.tokenizer = self._setup_model_and_tokenizer(
+                self.model,
+                self.tokenizer,
+            )
         else:
-            if self.tokenizer.chat_template is None:
-                self.model, self.tokenizer = self.setup_chat_format(self.model, self.tokenizer)
+            logger.error("Tokenizer is not initialized")
 
         try:
-
-            input_ids = self.tokenizer.apply_chat_template(inputs, tokenize=True, add_generation_prompt=True, return_tensors="pt")
+            input_ids = self.tokenizer.apply_chat_template(
+                inputs,
+                tokenize=True,
+                add_generation_prompt=True,
+                return_tensors="pt",
+            )
             # Generate response using the model
             outputs = self.model.generate(
                 input_ids.to(self.device),
@@ -361,7 +356,7 @@ def load_tokenizer(
                     f"'{pretrained_model_name_or_path}'"
                     f" from '{local_model_path}'",
                 )
-            self.tokenizer.padding_side = 'right'
+            self.tokenizer.padding_side = "right"
 
         except Exception as e:
             # Handle exceptions during model loading,
@@ -374,7 +369,7 @@ def load_tokenizer(
             logger.error(error_message)
 
             raise
-    
+
     def setup_chat_format(
         self,
         model: PreTrainedModel,
@@ -382,21 +377,34 @@ def setup_chat_format(
         resize_to_multiple_of: Optional[int] = None,
     ) -> Tuple[PreTrainedModel, PreTrainedTokenizer]:
         """
-        Setup chat format by adding special tokens to the tokenizer, setting the correct format, and extending the embedding layer of the model based on the new special tokens.
+        Setup chat format by adding special tokens to the tokenizer,
+        setting the correct format, and extending the embedding layer
+        of the model based on the new special tokens.
 
         Args:
         model (`~transformers.PreTrainedModel`): The model to be modified.
-        tokenizer (`~transformers.PreTrainedTokenizer`): The tokenizer to be modified.
-        format (`Optional[Literal["chatml"]]`): The format to be set. Defaults to "chatml".
-        resize_to_multiple_of (`Optional[int]`): Number to resize the embedding layer to. Defaults to None.
+        tokenizer (`~transformers.PreTrainedTokenizer`): The tokenizer
+                                                         to be modified.
+        format (`Optional[Literal["chatml"]]`): The format to be set.
+                                                Defaults to "chatml".
+        resize_to_multiple_of (`Optional[int]`): Number to resize
+                                                 the embedding layer to.
+                                                 Defaults to None.
         Returns:
-        model (`~transformers.PreTrainedModel`): The modified model.
-        tokenizer (`~transformers.PreTrainedTokenizer`): The modified tokenizer.
+        model (`~transformers.PreTrainedModel`): modified model.
+        tokenizer (`~transformers.PreTrainedTokenizer`): modified tokenizer.
         """
-        
 
         # set special tokens and them
-        tokenizer.add_special_tokens({"additional_special_tokens": ['<|system|>', '<|user|>', '<|assistant|>']})
+        tokenizer.add_special_tokens(
+            {
+                "additional_special_tokens": [
+                    "<|system|>",
+                    "<|user|>",
+                    "<|assistant|>",
+                ],
+            },
+        )
         # set chat format for tokenizer
         tokenizer.chat_template = """{% for message in messages %}
         {% if message['role'] == 'user' %}
@@ -411,9 +419,13 @@ def setup_chat_format(
         {% endif %}
         {% endfor %}"""
 
-        # resize embedding layer to a multiple of 64, https://x.com/karpathy/status/1621578354024677377
+        # resize embedding layer to a multiple of 64,
+        # https://x.com/karpathy/status/1621578354024677377
         model.resize_token_embeddings(
-            len(tokenizer), pad_to_multiple_of=resize_to_multiple_of if resize_to_multiple_of is not None else None
+            len(tokenizer),
+            pad_to_multiple_of=resize_to_multiple_of
+            if resize_to_multiple_of is not None
+            else None,
         )
         # Update the model config to use the new eos & bos tokens
         if getattr(model, "config", None) is not None:
@@ -428,6 +440,53 @@ def setup_chat_format(
 
         return model, tokenizer
 
+    def _setup_model_and_tokenizer(
+        self,
+        model: PreTrainedModel,
+        tokenizer: PreTrainedTokenizer,
+    ) -> Tuple[PreTrainedModel, PreTrainedTokenizer]:
+        """
+        Set up the model and tokenizer based on the
+        local model path and chat template.
+
+        This method checks if a local model path exists
+        and if the tokenizer has a chat template.
+        It then sets up the chat format if necessary
+        and loads pre-trained token embeddings.
+
+        Args:
+            model (PreTrainedModel): The pre-trained model to set up.
+            tokenizer (PreTrainedTokenizer): The tokenizer
+                                             associated with the model.
+
+        Returns:
+            Tuple[PreTrainedModel, PreTrainedTokenizer]:
+            The potentially modified model and tokenizer.
+
+        Note:
+            This method modifies the model and tokenizer
+            in-place but also returns them for convenience.
+        """
+        if self.local_model_path is not None:
+            if tokenizer.chat_template is not None:
+                if model.get_input_embeddings().weight.size()[0] != len(
+                    tokenizer,
+                ):
+                    model, tokenizer = self.setup_chat_format(model, tokenizer)
+                    loaded_embed_tokens_weights = torch.load(
+                        f"{self.local_model_path}/embed_tokens_weights.pt",
+                    )
+                    model.get_input_embeddings().weight.data.copy_(
+                        loaded_embed_tokens_weights,
+                    )
+                    del loaded_embed_tokens_weights
+                    torch.cuda.empty_cache()
+            elif tokenizer.chat_template is None:
+                model, tokenizer = self.setup_chat_format(model, tokenizer)
+        elif tokenizer.chat_template is None:
+            model, tokenizer = self.setup_chat_format(model, tokenizer)
+
+        return model, tokenizer
 
     def fine_tune(
         self,
@@ -471,6 +530,44 @@ def fine_tune(
             )
             raise
 
+    # Function to reformat a single row
+    def _reformat_row(self, row: Dict[str, List[str]]) -> List[Dict[str, str]]:
+        """
+        Reformat a single row of conversation data
+        into a list of message dictionaries.
+
+        This method takes a row from the dataset,
+        which contains a list of conversation
+        turns, and reformats it into a list of dictionaries.
+        Each dictionary represents
+        a message in the conversation, with 'role' and 'content' keys.
+
+        Args:
+            row (Dict[str, List[str]]): A dictionary containing a
+                                        'conversations' key
+                                        with a list of two strings:
+                                        the user's input
+                                        and the assistant's response.
+
+        Returns:
+            List[Dict[str, str]]: A list of three dictionaries
+                                  representing the system message,
+                                  user message, and assistant message.
+
+        Example:
+            Input row: {"conversations": ["User input", "Assistant response"]}
+            Output: [
+                {"role": "system", "content": "You're a helpful assistant."},
+                {"role": "user", "content": "User input"},
+                {"role": "assistant", "content": "Assistant response"}
+            ]
+        """
+        return [
+            {"role": "system", "content": "You're a helpful assistant."},
+            {"role": "user", "content": row["conversations"][0]},
+            {"role": "assistant", "content": row["conversations"][1]},
+        ]
+
     def _formatting_prompts_func(
         self,
         example: Dict[str, List[List[str]]],
@@ -543,39 +640,20 @@ def fine_tune_training(
             as part of the log/model fodler name.
         """
 
-        if self.local_model_path is not None:
-            if tokenizer.chat_template is not None:
-                if model.get_input_embeddings().weight.size()[0] != len(tokenizer):
-                    model, tokenizer = self.setup_chat_format(model, tokenizer)
-                    loaded_embed_tokens_weights = torch.load(self.local_model_path+'/embed_tokens_weights.pt')
-                    model.get_input_embeddings().weight.data.copy_(loaded_embed_tokens_weights)
-                    del loaded_embed_tokens_weights
-                    torch.cuda.empty_cache()
-            else:
-                if tokenizer.chat_template is None:
-                    model, tokenizer = self.setup_chat_format(model, tokenizer)
-        else:
-            if tokenizer.chat_template is None:
-                model, tokenizer = self.setup_chat_format(model, tokenizer)
+        model, tokenizer = self._setup_model_and_tokenizer(model, tokenizer)
 
         dataset = load_dataset(data_path, split="train", token=token)
 
         # # filter out input sequences that are longer than certain threshold
         dataset_reduced = dataset.filter(
             lambda x: len(x["conversations"][0] + x["conversations"][1])
-            <=3000,
+            <= 3000,
         )
 
-        # Function to reformat a single row
-        def reformat_row(row):
-            return [
-                    {"role": "system", "content": "You're a helpful assistant."},
-                    {"role": "user", "content": row['conversations'][0]},
-                    {"role": "assistant", "content": row['conversations'][1]}
-            ]
-
         # Apply the reformatting function to each row in the dataset
-        formatted_dataset = [reformat_row(row) for row in dataset_reduced]
+        formatted_dataset = [
+            self._reformat_row(row) for row in dataset_reduced
+        ]
         formatted_dataset = Dataset.from_dict({"messages": formatted_dataset})
 
         training_defaults = {
@@ -600,12 +678,14 @@ def reformat_row(row):
                 )
 
         if fine_tune_config.get("continue_lora_finetuning") is True:
-                self.lora_config = PeftConfig.from_pretrained(self.local_model_path)
-                model = PeftModel.from_pretrained(model, self.local_model_path)
-                # unfreeze lora parameters. Assuming 'lora' is in the layer name.
-                for name, param in model.named_parameters(): 
-                    if "lora" in name:
-                        param.requires_grad = True
+            self.lora_config = PeftConfig.from_pretrained(
+                self.local_model_path,
+            )
+            model = PeftModel.from_pretrained(model, self.local_model_path)
+            # unfreeze lora parameters. Assuming 'lora' is in the layer name.
+            for name, param in model.named_parameters():
+                if "lora" in name:
+                    param.requires_grad = True
         else:
             if lora_config_default:
                 self.lora_config = LoraConfig(**lora_config_default)
@@ -614,14 +694,8 @@ def reformat_row(row):
         if output_dir is not None:
             training_defaults["output_dir"] = output_dir
 
-        collator = DataCollatorForCompletionOnlyLM(
-            response_template=" ### Answer:",
-            tokenizer=tokenizer,
-        )
-
         trainer_args = transformers.TrainingArguments(**training_defaults)
 
-        
         trainer = SFTTrainer(
             model=model,
             tokenizer=tokenizer,
@@ -646,6 +720,7 @@ def reformat_row(row):
             trainer.train()
         except Exception as e:
             import traceback
+
             logger.error(f"Error during training: {e}")
             traceback.print_exc()
             raise
@@ -653,7 +728,6 @@ def reformat_row(row):
         now = datetime.now()
         time_string = now.strftime("%Y-%m-%d_%H-%M-%S")
 
-
         # save model
         model_name = (
             f"sft_{model.config.name_or_path.split('/')[-1]}_{time_string}"
@@ -664,12 +738,21 @@ def reformat_row(row):
             model_path = os.path.join(os.path.dirname(__file__), model_name)
         trainer.save_model(model_path)
 
-        # save token embeddings because it is resized due to the addition of new special tokens
-        embed_tokens_weights = model.get_input_embeddings().weight.data.clone().detach()
-        torch.save(embed_tokens_weights, model_path+'/embed_tokens_weights.pt')
+        # save token embeddings because it is resized
+        # due to the addition of new special tokens
+        embed_tokens_weights = (
+            model.get_input_embeddings().weight.data.clone().detach()
+        )
+        torch.save(
+            embed_tokens_weights,
+            model_path + "/embed_tokens_weights.pt",
+        )
 
-        with open(os.path.join(model_path, "log_history.json"), "w", encoding="utf-8") as f:
+        with open(
+            os.path.join(model_path, "log_history.json"),
+            "w",
+            encoding="utf-8",
+        ) as f:
             json.dump(trainer.state.log_history, f)
 
         return model
-

From 06d3ae147bd791a72eb03e99858a3599e56d5aac Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 17 Jul 2024 22:25:04 +0800
Subject: [PATCH 53/55] when `continue_lora_finetuning ` is `True`, check if
 model is already `PeftModel` before convert it to `PeftModel`

---
 .../README.md                                 | 25 +++++++++----------
 .../huggingface_model.py                      | 22 ++++++++--------
 2 files changed, 24 insertions(+), 23 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
index 99e3fcaa7..727e7d765 100644
--- a/examples/conversation_with_agent_with_finetuned_model/README.md
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -27,30 +27,29 @@ When initializing an agent, the following parameters need specification:
 ```
 System:
 
-Explain in simple terms how the attention mechanism of a transformer model works.
+You're a helpful assistant.
+
+User:
+
+Who are you?
 
 Assistant:
 
-pessimal answer: Attn explications: Attn is a type of attention mechanism. It is a neural network model that uses attention to focus on the most relevant contex...
-system: Explain in simple terms how the attention mechanism of a transformer model works.
-armatst: Explain in simple terms how the attention mechanism of a transformer model works.
-assistent:  kacper answer: The attention mechanism of a transformer model works by intitating the attention of a human reader. It glances at the contex...
-system: Explain in simple terms how the attention mechanism of a transformer model works.
-assistent: Explain in simple terms how the
+I am a woman who is passionate about life, my family, my friends, my work, and my community. I am a woman who is a wife, a mother, a daughter, a sister, a friend, a teacher, a coach, a mentor, a leader, a volunteer, a writer, a reader, a traveler, a gardener, a cook, a baker, a cra
 ```
 ### After Fine-tuning (with the default configuration in `model_configs`):
 ```
 System:
 
-Explain in simple terms how the attention mechanism of a transformer model works.
+You're a helpful assistant.
+
+User:
+
+Who are you?
 
 Assistant:
 
-Sure, the attention mechanism of a transformer model is an important part of the model's ability to generate coherent text. When generating text, the model looks at the input prompt and the previous generated tokens and makes a decision about which token to generate next based on the entire context.
-Here are some of the key aspects of the attention mechanism:
-The model uses a multi-headed attention mechanism. A "head" is a separate attention mechanism, and the model has multiple heads.
-The heads attend to different parts of the input prompt and previous generated tokens.
-The heads output weights used in the final output layer to
+I am a language model trained by Google to answer questions.
 ```
 (This example is trained with the default setting, with training time 872 seconds and 9.914 GB gpu memory cost. Reduce training batch size can reduce the memory required. Note that the model is loaded in 4 bits (i.e., QLoRA)).
 
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index c3b0e0b01..d8bf7791d 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -86,6 +86,7 @@ def __init__(
         self.max_length = max_length  # Set max_length as an attribute
         self.pretrained_model_name_or_path = pretrained_model_name_or_path
         self.local_model_path = local_model_path
+        self.lora_config = None
         script_path = os.path.abspath(__file__)
         script_dir = os.path.dirname(script_path)
         dotenv_path = os.path.join(script_dir, ".env")
@@ -660,13 +661,12 @@ def fine_tune_training(
             "per_device_train_batch_size": 4,
             "gradient_accumulation_steps": 1,
             "gradient_checkpointing": False,
-            "num_train_epochs": 3,
+            "num_train_epochs": 1,
             "output_dir": "./",
             "optim": "paged_adamw_8bit",
             "logging_steps": 1,
         }
 
-        self.lora_config = None
         lora_config_default = {}
 
         if fine_tune_config is not None:
@@ -678,14 +678,16 @@ def fine_tune_training(
                 )
 
         if fine_tune_config.get("continue_lora_finetuning") is True:
-            self.lora_config = PeftConfig.from_pretrained(
-                self.local_model_path,
-            )
-            model = PeftModel.from_pretrained(model, self.local_model_path)
-            # unfreeze lora parameters. Assuming 'lora' is in the layer name.
-            for name, param in model.named_parameters():
-                if "lora" in name:
-                    param.requires_grad = True
+            if not isinstance(model, PeftModel):
+                self.lora_config = PeftConfig.from_pretrained(
+                    self.local_model_path,
+                )
+                model = PeftModel.from_pretrained(model, self.local_model_path)
+                # unfreeze lora parameters. Assuming
+                # 'lora' is in the lora layer name.
+                for name, param in model.named_parameters():
+                    if "lora" in name:
+                        param.requires_grad = True
         else:
             if lora_config_default:
                 self.lora_config = LoraConfig(**lora_config_default)

From 50f7ae7736d235947476ca6bc3d28becda242eff Mon Sep 17 00:00:00 2001
From: zyzhang1130 <36942574+zyzhang1130@users.noreply.github.com>
Date: Wed, 17 Jul 2024 22:42:24 +0800
Subject: [PATCH 54/55] update README regarding ``continue_lora_finetuning` and
 `lora_config` parameters

---
 examples/conversation_with_agent_with_finetuned_model/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/README.md b/examples/conversation_with_agent_with_finetuned_model/README.md
index 727e7d765..76f741ac3 100644
--- a/examples/conversation_with_agent_with_finetuned_model/README.md
+++ b/examples/conversation_with_agent_with_finetuned_model/README.md
@@ -19,7 +19,7 @@ When initializing an agent, the following parameters need specification:
 - `local_model_path` (str): Local path to the model (defaults to loading from Hugging Face if not provided).
 - `data_path` (str): Path to training data (fine-tuning is skipped if not provided).
 - `device` (str): The device (e.g., 'cuda', 'cpu') for model operation, defaulting to 'cuda' if available.
-- `fine_tune_config` (dict, Optional): A configuration dictionary for fine-tuning the model. It allows specifying hyperparameters and other training options that will be passed to the fine-tuning method. If not provided, default settings will be used. This allows for customization of the fine-tuning process to optimize model performance based on specific requirements.
+- `fine_tune_config` (dict, Optional): A configuration dictionary for fine-tuning the model. It allows specifying hyperparameters and other training options that will be passed to the fine-tuning method. If not provided, default settings will be used. This allows for customization of the fine-tuning process to optimize model performance based on specific requirements. Note that if `continue_lora_finetuning` is set to `True`, `lora_config` should not be specified since the previously saved peft model's config will be used instead. If `continue_lora_finetuning` is set to `False`, `lora_config` should be specified.
 - `huggingface_token` (from .env file): Token required for models needing authentication from Hugging Face.
 
 ## Example Ouputs

From 99cecb6a585820bade464b09ab8cd2037674e743 Mon Sep 17 00:00:00 2001
From: Ze Yu Zhang <zyzhang1130@gmail.com>
Date: Fri, 23 Aug 2024 19:43:17 +0800
Subject: [PATCH 55/55] removed # pylint: disable=useless-parent-delegation in
 FinetuneDialogAgent

---
 .../FinetuneDialogAgent.py                    |  2 +-
 ...rsation_with_agent_with_finetuned_model.py |  6 ++++-
 .../huggingface_model.py                      | 23 ++++++++++---------
 3 files changed, 18 insertions(+), 13 deletions(-)

diff --git a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
index f3167fde4..e03fa2423 100644
--- a/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
+++ b/examples/conversation_with_agent_with_finetuned_model/FinetuneDialogAgent.py
@@ -43,7 +43,6 @@ def __init__(
         Note:
             Refer to `class DialogAgent(AgentBase)` for more information.
         """
-        # pylint: disable=useless-parent-delegation
         super().__init__(
             name,
             sys_prompt,
@@ -51,6 +50,7 @@ def __init__(
             use_memory,
             memory_config,
         )
+        self.finetune = True
 
     def load_model(
         self,
diff --git a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
index 7e297aee9..292694905 100644
--- a/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/conversation_with_agent_with_finetuned_model.py
@@ -9,7 +9,10 @@
 Features include model and tokenizer loading,
 and fine-tuning on the lima dataset with adjustable parameters.
 """
-
+# This import is necessary for AgentScope to properly use
+# HuggingFaceWrapper even though it's not explicitly used in this file.
+# To remove the pylint disable without causing issues
+# HuggingFaceWrapper needs to be put under src/agentscope/agents.
 # pylint: disable=unused-import
 from huggingface_model import HuggingFaceWrapper
 from FinetuneDialogAgent import FinetuneDialogAgent
@@ -52,6 +55,7 @@ def main() -> None:
                 # loading a model saved as lora model
                 "fine_tune_config": {
                     "continue_lora_finetuning": False,
+                    "max_seq_length": 4096,
                     "lora_config": {
                         "r": 16,
                         "lora_alpha": 32,
diff --git a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
index d8bf7791d..9e0731e25 100644
--- a/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
+++ b/examples/conversation_with_agent_with_finetuned_model/huggingface_model.py
@@ -80,10 +80,13 @@ def __init__(
                                                fine-tuning the model.
             **kwargs: Additional keyword arguments.
         """
-        super().__init__(config_name=config_name)
+        super().__init__(
+            config_name=config_name,
+            model_name=pretrained_model_name_or_path,
+        )
         self.model = None
         self.tokenizer = None
-        self.max_length = max_length  # Set max_length as an attribute
+        self.max_length = max_length
         self.pretrained_model_name_or_path = pretrained_model_name_or_path
         self.local_model_path = local_model_path
         self.lora_config = None
@@ -358,6 +361,8 @@ def load_tokenizer(
                     f" from '{local_model_path}'",
                 )
             self.tokenizer.padding_side = "right"
+            if self.tokenizer.pad_token is None:
+                self.tokenizer.pad_token = self.tokenizer.eos_token
 
         except Exception as e:
             # Handle exceptions during model loading,
@@ -666,10 +671,13 @@ def fine_tune_training(
             "optim": "paged_adamw_8bit",
             "logging_steps": 1,
         }
+        max_seq_length_default = 4096
 
         lora_config_default = {}
 
         if fine_tune_config is not None:
+            if fine_tune_config.get("max_seq_length") is not None:
+                max_seq_length_default = fine_tune_config["max_seq_length"]
             if fine_tune_config.get("training_args") is not None:
                 training_defaults.update(fine_tune_config["training_args"])
             if fine_tune_config.get("lora_config") is not None:
@@ -709,7 +717,7 @@ def fine_tune_training(
                 else {}
             ),
             args=trainer_args,
-            max_seq_length=4096,
+            max_seq_length=max_seq_length_default,
         )
 
         logger.info(
@@ -718,14 +726,7 @@ def fine_tune_training(
             timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
         )
 
-        try:
-            trainer.train()
-        except Exception as e:
-            import traceback
-
-            logger.error(f"Error during training: {e}")
-            traceback.print_exc()
-            raise
+        trainer.train()
 
         now = datetime.now()
         time_string = now.strftime("%Y-%m-%d_%H-%M-%S")