| # Flash Attention - CUDA 12, PyTorch 2.6, Python 3.10 |
| flash-attn @ https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.3/flash_attn-2.7.3+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl |
|
|
| # Core ML/AI Libraries |
| torch==2.6.0 |
| torchvision |
| accelerate>=0.24.0 |
|
|
| # Transformers - using version compatible with both sets of models |
| transformers==4.57.1 |
| tokenizers>=0.20.3 |
| transformers-stream-generator |
|
|
| # Hugging Face |
| huggingface_hub |
| hf_xet |
| spaces>=0.20.0 |
|
|
| # Model architectures and utilities |
| xformers |
| einops |
| peft |
|
|
| # Vision & Image Processing |
| qwen-vl-utils |
| albumentations |
| opencv-python |
| pyvips |
| pyvips-binary |
| Pillow>=10.0.0 |
| timm |
| supervision |
|
|
| # Document Processing |
| docling-core |
| python-docx |
| pymupdf |
| pdf2image |
| markdown |
| html2text |
|
|
| # PDF Generation |
| reportlab |
| fpdf |
|
|
| # Text Processing |
| sentencepiece>=0.1.99 |
| num2words |
| protobuf>=3.20.0 |
|
|
| # Utilities |
| loguru |
| requests |
| httpx |
| click |
| addict |
| easydict |
| safetensors>=0.4.0 |
|
|
| # Web Interface |
| gradio>=4.0.0 |
|
|
| # Video Processing |
| av |
