delete and build_stream_transfer_data

Author: lizexu123

fastdeploy/model_executor/pre_and_post_process.py

@@ -15,7 +15,7 @@
 """
 
 import queue
-from typing import Dict, Optional, Union
+from typing import Dict, List, Optional, Union
 
 import numpy as np
 import paddle
@@ -85,7 +85,7 @@
         speculate_limit_thinking_content_length_v2,
     )
 
-from fastdeploy.output.pooler import PoolerOutput
+from fastdeploy.output.pooler import PoolerOutput, PoolingSequenceGroupOutput
 from fastdeploy.output.stream_transfer_data import DecoderState, StreamTransferData
 from fastdeploy.worker.output import ModelOutputData, ModelRunnerOutput, SamplerOutput
 
@@ -239,7 +239,7 @@ def pre_process(
     )
 
 
-def _build_stream_transfer_data(output_tokens: np.ndarray, pooler_outputs: None):
+def _build_stream_transfer_data(output_tokens: np.ndarray, pooler_outputs: List[PoolingSequenceGroupOutput] = None):
     """Split output_tokens and output"""
 
     stream_transfer_datas = []

fastdeploy/worker/gpu_model_runner.py

@@ -1536,12 +1536,6 @@ def _dummy_pooler_run_task(
         assert len(num_scheduled_tokens_list) == num_reqs
 
         req_num_tokens = num_tokens // num_reqs
-
-        print("num_tokens", num_tokens)
-        print("max_num_seqs", max_num_seqs)
-        print("num_reqs", num_reqs)
-        print("min_tokens_per_req", min_tokens_per_req)
-        print("num_scheduled_token_list", num_scheduled_tokens_list)
         dummy_prompt_lens = paddle.to_tensor(num_scheduled_tokens_list, dtype="int64")
         dummy_token_ids = paddle.zeros(
             [num_reqs, req_num_tokens],