| cd /nfs/ywang29/video_reward/DanceGRPO | |
| source activate | |
| # conda create -n wan python==3.10 -y | |
| # conda activate wan | |
| ./env_setup.sh fastvideo | |
| pip3 install moviepy | |
| pip3 install protobuf==3.20.0 | |
| # mkdir videos | |
| pip3 install huggingface_hub==0.24.0 | |
| pip3 install tf-keras | |
| pip3 install trl | |
| # pip3 install transformers==4.51.0 | |
| pip install torchvision==0.21.0 | |
| pip install pydantic -U | |
| pip install qwen_vl_utils | |
| pip install librosa | |
| pip install wandb | |
| cd /nfs/ywang29/Reward_finetuning/VideoX-Fun | |
| export LD_LIBRARY_PATH=/usr/local/cuda/lib64 | |
| # pip install -U accelerator | |
| pip install transformers==4.57 | |
| pip install accelerate -U | |
| pip install flash-attn==2.5.8 --no-build-isolation | |
| # cp /nfs/ywang29/Reward_finetuning/VideoX-Fun/scripts/wan2.1/modeling_qwen2_5_vl.py /opt/conda/lib/python3.10/site-packages/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py | |