Search results

feat: W8A8 Accurate GEMV kernel (<0.5% error)

Status: Open.

#123 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 28, 2025

refactor: Reorganize kernel directory structure for clarity

Status: Open.

#122 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 28, 2025

feat: Auto-tuning kernel dispatcher with size-aware selection

enhancementNew feature or requestNew feature or request

Status: Open.

#120 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 26, 2025

feat: Auto-quantization based on available VRAM

enhancementNew feature or requestNew feature or request

Status: Open.

#119 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 26, 2025

[v0.2.17]feat: FP8 model loading and inference support

enhancementNew feature or requestNew feature or request

Status: Open.

#118 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 26, 2025

[v0.2.17] feat(triton): Triton Backend MVP

Status: Open.

#116 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 25, 2025

feat(model): QAT/QAD/Pruned Model Loading Support

Status: Open.

#115 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 25, 2025

feat(decode): Medusa Multi-Head Speculative Decoding

Status: Open.

#114 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 25, 2025

feat(decode): External Draft Model Speculative Decoding

Status: Open.

#113 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 25, 2025

feat(quantize): Post-Training Quantization (PTQ) - INT4/AWQ/GPTQ support

Status: Open.

#112 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 25, 2025

CUTLASS SM120 FP8 GEMM: TMA descriptor alignment fix (misaligned address resolved)

Status: Open.

#107 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 23, 2025

feat: SM120 (RTX 5090) support via FP8/FP6/FP4 precision

Status: Open.

#91 In m96-chan/PyGPUkit;

· m96-chan opened on Dec 16, 2025