Search results

Sparse Decode Mla kernel opt plan On blcakwell

Status: Open.

#120 In deepseek-ai/FlashMLA;

· lw921014 opened on Nov 9, 2025

Built FlashMLA for Windows and Nvidia sm_120 (workstation/50s) cards

Status: Open.

#119 In deepseek-ai/FlashMLA;

· IISuperluminaLII opened on Oct 30, 2025

Why is it much slower than Tilelang?

Status: Open.

#118 In deepseek-ai/FlashMLA;

· zzyplaybasketball opened on Oct 27, 2025

how to test sparse attention in test_flash_mla_decoding

Status: Open.

#116 In deepseek-ai/FlashMLA;

· jerry-dream-fu opened on Oct 24, 2025

No code for sm100 dense mla.

Status: Open.

#113 In deepseek-ai/FlashMLA;

· engineer1109 opened on Oct 17, 2025

RuntimeError: Sparse BF16 MLA is not supported on SM90

Status: Open.

#110 In deepseek-ai/FlashMLA;

· hermosayhl opened on Oct 14, 2025

metadata api compatibility issue

Status: Open.

#108 In deepseek-ai/FlashMLA;

· mingyangHao opened on Oct 13, 2025

Missing sparse backward kernel.

Status: Open.

#107 In deepseek-ai/FlashMLA;

· EricLina opened on Oct 9, 2025

Kronecker factors instead of LoRA?

Status: Open.

#102 In deepseek-ai/FlashMLA;

· bghira opened on Sep 29, 2025

Char datatype assumption

Status: Open.

#93 In deepseek-ai/FlashMLA;

· green-br opened on Sep 4, 2025

Partner Listing Request：PaddlePaddle 的 PyTorch C++/Python 接口兼容项目

Status: Open.

#86 In deepseek-ai/FlashMLA;

· yongqiangma opened on Aug 25, 2025

Why block_size_n equals page_block_size

Status: Open.

#84 In deepseek-ai/FlashMLA;

· mingyangHao opened on Aug 14, 2025