- Notifications
You must be signed in to change notification settings - Fork 0
All issues
Issue creation is restricted in this repository
Issues
is:issuestate:open
is:issue state:open
Search results
FA4: Flash Attention 4 with WGMMA for SM120
enhancementNew feature or requestNew feature or requestStatus: Open.#192In m96-chan/PyGPUkit;- Status: Open.#191In m96-chan/PyGPUkit;
- Status: Open.#190In m96-chan/PyGPUkit;
perf(diffusion): FLUX.1 transformer performance optimization
enhancementNew feature or requestNew feature or requestStatus: Open.#187In m96-chan/PyGPUkit;- Status: Open.#186In m96-chan/PyGPUkit;
- Status: Open.#184In m96-chan/PyGPUkit;
perf(tts): Kokoro layers use CPU fallback (numpy) instead of GPU kernels
enhancementNew feature or requestNew feature or requestv0.3Advanced: Triton backend, advanced opsAdvanced: Triton backend, advanced opsStatus: Open.#182In m96-chan/PyGPUkit;perf(llm): Batch decode missing CUDA Graph and zero-alloc optimizations
enhancementNew feature or requestNew feature or requestv0.2Scheduler: memory + bandwidth guaranteesScheduler: memory + bandwidth guaranteesStatus: Open.#181In m96-chan/PyGPUkit;perf(asr): Whisper conv1d uses CPU fallback instead of GPU kernel
enhancementNew feature or requestNew feature or requestv0.3Advanced: Triton backend, advanced opsAdvanced: Triton backend, advanced opsStatus: Open.#180In m96-chan/PyGPUkit;RFC: Universal Model Architecture Support (GPT/MoE/VLM/LRM/LAM/HLM/LCM)
enhancementNew feature or requestNew feature or requestv0.3Advanced: Triton backend, advanced opsAdvanced: Triton backend, advanced opsStatus: Open.#176In m96-chan/PyGPUkit;- Status: Open.#172In m96-chan/PyGPUkit;
- Status: Open.#170In m96-chan/PyGPUkit;