Search results

Can not install DeepGEMM with commit of 95e81b3

Status: Open.

#127 In deepseek-ai/DeepGEMM;

· lxbworld opened on Jul 11, 2025

TMA selection of get_best_configs in gemm.py

Status: Open.

#126 In deepseek-ai/DeepGEMM;

· joyeamd opened on Jul 10, 2025

[Question] Do you have any plans to improve grouped wgrad gemm?

Status: Open.

#125 In deepseek-ai/DeepGEMM;

· shcho1118 opened on Jul 9, 2025

[Question] How to call DeepGemm kernel from C++

Status: Open.

#123 In deepseek-ai/DeepGEMM;

· WanchaoYao opened on Jul 4, 2025

The reduction of GPU core frequency leads to significant fluctuations in GEMM performance

Status: Open.

#120 In deepseek-ai/DeepGEMM;

· yaning223 opened on Jun 25, 2025

Will swap ab tech improve small m (decode phase) performance?

Status: Open.

#114 In deepseek-ai/DeepGEMM;

· hengxinCheung opened on Jun 17, 2025

Wgrad gemm kernel illegal memory access with (m, n, k) in (512, 1024, 147456)

Status: Open.

#110 In deepseek-ai/DeepGEMM;

· RuiWang1998 opened on Jun 10, 2025

[Question] The clamp value in per_token_cast_to_fp8

Status: Open.

#109 In deepseek-ai/DeepGEMM;

· chenyu-jiang opened on Jun 9, 2025

从训练的角度出发，这个可以用在megatron里面吗

Status: Open.

#108 In deepseek-ai/DeepGEMM;

· duzhixing opened on Jun 4, 2025

[Help] Starter script for nn.Linear for fwd,bwd

Status: Open.

#102 In deepseek-ai/DeepGEMM;

· ajWithNucleus opened on May 17, 2025

m_grouped_gemm_fp8_fp8_bf16_nt_contiguous's requiresment of 128 padding for each expert causes computation waste

Status: Open.

#98 In deepseek-ai/DeepGEMM;

· xutizhou opened on May 14, 2025

fp8_wgrad_gemm Support unaligned n,k and gmem stride?

Status: Open.

#96 In deepseek-ai/DeepGEMM;

· John-tsy opened on May 12, 2025