Spaces:
Running on Zero
Running on Zero
Commit History
3 days ago
-
Fix OOM and extreme lag by bypassing unused reference attention in single-human mode and chunking multi-human reference attention Running
Daniel03a50fa -
Accelerate flash_attention using xformers Cutlass and contiguous SDPA kernels
Daniel23defbb -
Optimize ZeroGPU duration to 300s, set recommended 15 steps, and remove torch.compile lag
Danield05af4e -
Disable layer-by-layer VRAM offloading to run DiT natively on GPU
Daniel5045d7c -
Fix tensor device mismatch between cuda and cpu in SDPA fallback
Daniel85175b0
4 days ago
-
Refine SDPA attention masking and adjust ZeroGPU duration
Daniel63e3991 -
Implement safe_memory_efficient_attention with PyTorch SDPA fallback
Verified signature5edc824 -
Disable empty_cache during diffusion loop to prevent dynamo handler collision
Verified signaturea8ab186 -
Add PyTorch native SDPA fallback in flash_attention
Verified signature67b2c08 -
Direct layer hidden_states collection in Wav2Vec2
Verified signature5547fec -
Set attn_implementation to eager for Wav2Vec2
Verified signatureb7c2342 -
Fix tokenizer path fallback to HF Hub model ids
Verified signatureca5c28f -
Fix utils init import get_gpu_manager
Verified signature5e918f1 -
Fix CUDA re-init in forked subprocess: defer model imports & lazy xformers
Verified signaturee579ffb -
Upgrade Gradio SDK to 6.14.0 and patch schema validation
Verified signature1dc8a76 -
Fix ZeroGPU compatibility: mock gcnArchName and decouple xfuser
Verified signaturefcb6a26 -
Deploy InfiniteTalk Gradio Demo with ZeroGPU support
Verified signature13c793c -
initial commit
Verified signature70af39c