Add Windows + RTX 5090 (Blackwell) support#70
Open
HurtzDonutStudios wants to merge 1 commit intoVAST-AI-Research:mainfrom
Open
Add Windows + RTX 5090 (Blackwell) support#70HurtzDonutStudios wants to merge 1 commit intoVAST-AI-Research:mainfrom
HurtzDonutStudios wants to merge 1 commit intoVAST-AI-Research:mainfrom
Conversation
Enables UniRig to run end-to-end on Windows with modern NVIDIA GPUs (tested RTX 5090 + CUDA 12.8 + PyTorch 2.11). Key changes: - CUDA DLL path registration for spconv/cumm on Windows - PyTorch 2.11 weights_only=False patch for checkpoint loading - torch_scatter/torch_cluster CPU fallback for missing CUDA kernels - flash_attn SDPA shim (PyTorch native attention, no CUDA compilation needed) - PTv3 non-flash attention path (enable_flash: False) - Shell script fixes: venv auto-activation, backslash path conversion - NPZ export fix for user_mode pipeline (skeleton → skin data chain) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
flash_attnSDPA shim so no CUDA flash-attention compilation is neededChanges
run.pyweights_only=Falserun.pytorch_scatter.segment_csrCPU fallbackrun.pyfps()CPU fallbacksal_perceiver.pyflash_attn_shim/,install_flash_attn_shim.pyenable_flash: Falsefor PTv3unirig_skin.yamleagerattention for OPT LLMunirig_ar_350m_*.yamlar.pylaunch/inference/*.shskin.pyTest Environment
Test plan
🤖 Generated with Claude Code