Skip to content

README.md

Latest commit

 

History

History
52 lines (37 loc) · 1.33 KB

README.md

File metadata and controls

52 lines (37 loc) · 1.33 KB

💻 Environment & Installation

To setup the training environment, run:

cd LongRM
pip install -r requirements.txt
# install flash attention
Download the suitable version of flash_attn from https://github.com/Dao-AILab/flash-attention/releases
pip install <path_to_flash_attn_whl_file>
pip install ring_flash_attn

🔥 Train

For Generative Model

To run the first training process:

bash scripts/sft.sh

To run the second training process:

bash bash scripts/simpo_grm.sh

For Discriminative Model

Directly run the second training process:

bash bash scripts/simpo_disrm.sh

📊 Evaluate

  • We provide the benchmark dataset and trained models in our Modelscope.

For Generative Model

modelscope download LCM_group/LongReward_Qwen3-8B --repo-type model --local_dir ./LongReward_Qwen3-8B

python evaluate/eval.py --model-path ./LongReward_Qwen3-8B --data-path ./LongReward-Bench

For Discriminative Model

modelscope download LCM_group/LongReward_Skywork-Reward-V2-Llama-3.1-8B --repo-type model --local_dir ./LongReward_Skywork-Reward-V2-Llama-3.1-8B

python evaluate/eval.py --model-path ./LongReward_Skywork-Reward-V2-Llama-3.1-8B --data-path ./LongReward-Bench --is-disrm