woctordho

triton-windows. Fork of the Triton language and compiler for Windows support and easy installation

SageAttention. Fork of SageAttention for Windows wheels and easy installation

transformers-qwen3-moe-fused. Fused Qwen3 MoE layer for faster training, compatible with Transformers, LoRA, bnb 4-bit quant, Unsloth. Also possible to train LoRA over GGUF

pkuholebackup.

ComfyUI-RadialAttn. RadialAttention in ComfyUI native workflow

typeset. 自动修正中文、英文、代码混合排版中的全半角、空格等问题

pkubbsbackup.

ACE-Step. Fork of ACE-Step v1.0 for LoRA training with < 10 GB VRAM

project-asteria. Project Asteria: A Naïve Introductory to Advanced Mathematics and Theoretical Physics for Gaokao Students

SpargeAttn. Fork of SpargeAttention (SparseSageAttention) for Windows wheels and easy installation

ComfyUI-FeatherOps. Fast fp16-fp8 mixed precision matmul on RDNA3/3.5 GPUs without native fp8

read-matrix. Reviews to The Matrix series

mathstudio-apk-modernized. MathStudio Android APK modernized for Android >= 7

sageattention-autotune. SageAttention with autotuned block sizes

evotensile. Evolutionary algorithm for hipBLASLt TensileLite config search

rdna35-isa-markdown. AMD GPU RDNA3.5 (Strix Halo) instruction set architecture manual in Markdown for AI retrieval

typeset-rs. 自动修正中文、英文、代码混合排版中的全半角、空格等问题