🎯 Brief Introduction

Phi-4-mini-instruct is a lightweight open model built upon synthetic data and filtered publicly available websites - with a focus on high-quality, reasoning dense data. The model belongs to the Phi-4 model family and supports 128K token context length. The model underwent an enhancement process, incorporating both supervised fine-tuning and direct preference optimization to support precise instruction adherence and robust safety measures.

Primary Use Cases

The model is intended for broad multilingual commercial and research use. The model provides uses for general purpose AI systems and applications which require:

• Memory/compute constrained environments
• Latency bound scenarios
• Strong reasoning (especially math and logic).

The model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features.

Model

• github
• huggingface

Usage for Training

Create a new file examples/train_lora/phi4_mini_lora_sft.yaml with the following content:

model_name_or_path: microsoft/Phi-4-mini-instruct

trust_remote_code: true

### method

stage: sft
do_train: true
finetuning_type: lora
lora_rank: 8
lora_target: all

### dataset

dataset: identity,alpaca_en_demo
template: phi4_mini

cutoff_len: 2048
max_samples: 1000
preprocessing_num_workers: 16
dataloader_num_workers: 4

### output

output_dir: saves/phi-4-mini-3.8B/lora/sft
logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: none  # choices: [none, wandb, tensorboard, swanlab, mlflow]

### train

per_device_train_batch_size: 1
gradient_accumulation_steps: 8
learning_rate: 1.0e-4
num_train_epochs: 3.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000
resume_from_checkpoint: null

### eval

# eval_dataset: alpaca_en_demo

# val_size: 0.1

# per_device_eval_batch_size: 1

# eval_strategy: steps

# eval_steps: 500

used

1 gpu

DISABLE_VERSION_CHECK=1 llamafactory-cli train examples/train_lora/phi4_mini_lora_sft.yaml
 1 node

DISABLE_VERSION_CHECK=1 FORCE_TORCHRUN=1 llamafactory-cli train examples/train_lora/phi4_mini_lora_sft.yaml

result

template: phi

template: phi4_mini

We conducted a comparison of the loss metrics between the phi template and phi4_mini template. The results indicate that the ​phi4_mini ​template is better suited for the training process of the current ​phi4_mini ​model (demonstrating better initial loss convergence and format compatibility with the target model.