Search results

[RFC] Support FP4 (E2M1) activation dispatch in MegaMoE

Status: Open.

#338 In deepseek-ai/DeepGEMM;

· wuudiz opened on May 19, 2026

Question about Nsight Compute results for sm100_fp8_fp4_mega_moe_impl on B300/Blackwell

Status: Open.

#336 In deepseek-ai/DeepGEMM;

· Stone749990226 opened on May 18, 2026

Will DeepGEMM support GEMM + 1d1d quantization all in one fused kernel?

Status: Open.

#335 In deepseek-ai/DeepGEMM;

· TheTinyTeddy opened on May 18, 2026

[Feature Request] Make binary artifacts portable on torch and Python versions

Status: Open.

#333 In deepseek-ai/DeepGEMM;

· Harry-Chen opened on May 12, 2026

get_paged_mqa_logits_metadata() does not work with large batch size due to excessive static shared memory impl

Status: Open.

#322 In deepseek-ai/DeepGEMM;

· gau-nernst opened on Apr 29, 2026

fix: operator precedence bug in pack_ue8m0_to_int assertion (mantissa check always passes)

Status: Open.

#309 In deepseek-ai/DeepGEMM;

· kuishou68 opened on Apr 19, 2026

[Bug] JIT cache lacks cross-process synchronization, can cause failures under multi-process parallelism

Status: Open.

#301 In deepseek-ai/DeepGEMM;

· Gregory-Pereira opened on Apr 11, 2026

CUDA kernel errors

Status: Open.

#300 In deepseek-ai/DeepGEMM;

· ljg-ieisystem opened on Apr 3, 2026

[RFC] Using tvm-ffi binding instead of python api to remove the dependency for python binding

Status: Open.

#297 In deepseek-ai/DeepGEMM;

· rainj-me opened on Mar 25, 2026

[Bug]: NVRTC JIT compilation fails on CUDA 12.8 (smxx_clean_logits.cuh: expression must have a constant value)

Status: Open.

#295 In deepseek-ai/DeepGEMM;

· LilySu opened on Mar 25, 2026

weight gradient kernels for dense and MoE backward 没有grad的反向算子

Status: Open.

#283 In deepseek-ai/DeepGEMM;

· ljg-ieisystem opened on Jan 23, 2026

[v3.2] deep_gemm::fp8_mqa_logits performs poorly on the H200

Status: Open.

#279 In deepseek-ai/DeepGEMM;

· thqq479 opened on Jan 16, 2026