FastH3-4step-Preview-VSA / requirements.txt
Mike0021's picture
exp8: wheel CUDA VSA active (einops fix)
257cf81 verified
Raw
History Blame Contribute Delete
1.04 kB
# MiniMax-H3 landed in diffusers 0.40.0 (modular pipelines only). Pinned because `h3_split_blocks.py` subclasses its
# block classes and `vsa_h3.py` patches `MiniMaxH3TransformerBlock` / `MiniMaxH3Attention` by name.
diffusers==0.40.0
# torch is deliberately unpinned: the ZeroGPU runtime preinstalls a supported build, and its bundled `pytorch-triton`
# is what compiles the vendored VSA block-sparse kernels. torchvision is not needed anywhere in this Space.
transformers
accelerate
# FastVideo's published CUDA kernels: the block-sparse VSA forward, sm_100a source with sm_120a cubins in the
# same fatbin (this pool's GPU). Only needs torch>=2.5, so the platform torch is untouched. If the op refuses
# to run, `vsa_h3` falls back to the vendored Triton kernels below.
fastvideo-kernel==0.3.4
# PyAV muxes the generated soundtrack onto the frames (`diffusers.utils.encode_video`).
av
pillow
numpy
safetensors>=0.8.0
# huggingface-hub is deliberately absent: it is platform-managed, and diffusers already constrains it to <2.
einops