Search results

Question About deepGEMM contributor and Collaboration

Status: Open.

#391 In deepseek-ai/DeepGEMM;

· DiegoCao opened on Jul 22, 2026

Origin & calibration methodology of the three cost-model constants in sm120.hpp

Status: Open.

#385 In deepseek-ai/DeepGEMM;

· json0701 opened on Jul 17, 2026

[DSv4] IS W4A16(mxfp4, act bf16) on the planning list?

Status: Open.

#374 In deepseek-ai/DeepGEMM;

· heroes999 opened on Jul 6, 2026

transform_sf_into_required_layout: no SM120/121 (arch_major=12) branch for (gran_mn=1, gran_k=32) NVFP4 expert scales -- blocks DeepSeek-V4 weight loading

Status: Open.

#372 In deepseek-ai/DeepGEMM;

· zbrad opened on Jul 6, 2026

Incompatibility between DeepEP V2 ElasticBuffer non-expanded output and DeepGEMM masked/contiguous grouped GEMM input format

Status: Open.

#370 In deepseek-ai/DeepGEMM;

· litiantian00 opened on Jul 1, 2026

question about tmem barrier

Status: Open.

#368 In deepseek-ai/DeepGEMM;

· wukong1992 opened on Jun 26, 2026

question about fp8fp4 mega moe weights load

Status: Open.

#367 In deepseek-ai/DeepGEMM;

· wukong1992 opened on Jun 25, 2026

[Feature] Will DeepGEMM support more advanced epilogue fusion for GEMM kernels?

Status: Open.

#366 In deepseek-ai/DeepGEMM;

· b8zhong opened on Jun 24, 2026

[BUG] Non-deterministic numerical problem of k_grouped_fp8_gemm_nt_contiguous

Status: Open.

#365 In deepseek-ai/DeepGEMM;

· FFGGSSJJ opened on Jun 23, 2026

[BUG] NVRTC build fails for sm90_tf32_hc_prenorm_gemm — cudaGridDependencySynchronize undefined (PDL intrinsic not available under NVRTC)

Status: Open.

#359 In deepseek-ai/DeepGEMM;

· zhehli688 opened on Jun 12, 2026

[Feature request] Documented clangd setup for CUDA kernel indexing

Status: Open.

#358 In deepseek-ai/DeepGEMM;

· Rachmanino opened on Jun 12, 2026

Does Mega MoE support RTX 50-series GPUs, and can it work in a PCIe-only setup without NVLink?

Status: Open.

#356 In deepseek-ai/DeepGEMM;

· wukong1992 opened on Jun 10, 2026