PSHuman

Running on Zero

PSHuman / mvdiffusion

2.5 MB

Ctrl+K

2 contributors

Implement attention slicing: chunk SDPA calls along the batch*heads dim since torch 2.8s fused kernels reject this shape on this GPU, forcing an O(seq_len^2 * batch) math fallback that OOMs even a 96GB card

63d710a verified 7 days ago

data
Migrated from GitHub almost 2 years ago
models_unclip
Implement attention slicing: chunk SDPA calls along the batch*heads dim since torch 2.8s fused kernels reject this shape on this GPU, forcing an O(seq_len^2 * batch) math fallback that OOMs even a 96GB card 7 days ago
pipelines
Migrated from GitHub almost 2 years ago