Spaces:
Running
on
Zero
Running
on
Zero
| # Model Training and Optimization | |
| transformers>=4.49.0 | |
| diffusers>=0.31.0 | |
| accelerate>=1.1.1 | |
| tokenizers>=0.20.3 | |
| flash-attn @ https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.4cxx11abiFALSE-cp310-cp310-linux_x86_64.whl | |
| # Configuration and Experiment Management | |
| hydra-core>=1.3.0 | |
| omegaconf>=2.3.0 | |
| pyyaml>=6.0 | |
| wandb>=0.15.0 | |
| # Video/Image Processing | |
| av==14.1.0 | |
| pillow>=9.5.0,<10.0.0 # Pinned for compatibility | |
| einops>=0.7.0 | |
| # Text Processing and Tokenization | |
| sentencepiece>=0.2.0 | |
| ftfy>=6.1.0 # For text cleaning | |
| huggingface-hub>=0.20.0 | |
| tqdm>=4.65.0 | |
| colorama>=0.4.6 | |
| click>=8.1.0 | |
| easydict>=1.10 | |
| msgpack>=1.0.5 # For message serialization | |
| pyzmq>=25.0.0 # For ZeroMQ (used in serving) | |
| gradio>=5.0.0 |