DELUXA

Systems-level contributor focused on PyTorch, AMD ROCm on Windows, and ML infrastructure.

Notable Contributions

Repository	Contribution
huggingface/transformers	Fix: Conditionally import `torch.distributed.fsdp` in `trainer_seq2seq.py`
Comfy-Org/ComfyUI	quant_ops: drop the dead ROCm triton arch gate
	Fix: Verify aotriton kernels actually launch before enabling pytorch attention
	Enable fp8 ops by default on gfx1200
pytorch/pytorch	[ROCm] Fix hipify path handling on Windows
	[elastic] Add Windows support for stdout/stderr redirects
invoke-ai/InvokeAI	feat(deps): ROCm 10 on Linux and Windows from AMD's index (#9643)
	feat(deps): install ROCm 10 from AMD's index on Linux and Windows
Dao-AILab/flash-attention	[ROCM] Fix windows issues (#2385)
triton-lang/triton	[AMD] Stop lowering bf16 multiply to v_dot2_bf16_bf16
vosen/ZLUDA	docs: clarify unofficial HIP SDK column refers to AMD nightlies
deepbeepmeep/WanGP	Refresh Windows AMD ROCm support for TheRock multi-arch wheels
	Update `AMD-INSTALLATION.md`
	Revise AMD installation guide
huggingface/accelerate	Fix: Conditionally import `torch.distributed.algorithms.join` in `accelerator.py`
LykosAI/StabilityMatrix	Fix: Pin OneTrainer ROCm bitsandbytes wheel to 0.49.1

---

_{Auto-updated via GitHub Actions}