Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Apr 2, 2025

• Feat: multi-node tensor parallel inference (#46)

  Show description for fa6a70d
  unix1986

  authored
  fa6a70d

  Copy full SHA for fa6a70d

Commits on Apr 1, 2025

• Support reasoning content (#48)

  Show description for 09d705f
  

  yabea

  and

  yangben

  authored
  09d705f

  Copy full SHA for 09d705f

Commits on Mar 24, 2025

• Merge pull request #44 from spetrel/deep_gemm

  Show description for 5b6e905
  unix1986

  authored
  5b6e905

  Copy full SHA for 5b6e905

• Integrate DeepGemm part1.

  spetrel

  committed
  e30a7b3

  Copy full SHA for e30a7b3

• Add fp8 block utils.

  spetrel

  committed
  fc201fe

  Copy full SHA for fc201fe

• Add DeepGemm library.

  spetrel

  committed
  626cd7f

  Copy full SHA for 626cd7f

• add deepseek r1 awq benchmark results

  Show description for 6ff12ff
  yabea

  authored
  6ff12ff

  Copy full SHA for 6ff12ff

• add deepseek r1 awq benchmark results

  

  yangben

  committed
  43cb7dc

  Copy full SHA for 43cb7dc

Commits on Mar 14, 2025

• Merge pull request #41 from spetrel/main

  Show description for 66b7f91
  unix1986

  authored
  66b7f91

  Copy full SHA for 66b7f91

• Integrate FlashMLA.

  spetrel

  committed
  abc4d5e

  Copy full SHA for abc4d5e

• Add libflash_mla.a

  spetrel

  committed
  2f0e9f7

  Copy full SHA for 2f0e9f7

• Update bmengine.

  spetrel

  committed
  20a9da0

  Copy full SHA for 20a9da0

Commits on Mar 10, 2025

• Update README.md

  unix1986

  authored
  beb286d

  Copy full SHA for beb286d

• Refactor multi_head_latent_attention.cpp

  spetrel

  committed
  1c0f9f7

  Copy full SHA for 1c0f9f7

• Change kv cache resizing step from 32 to 64.

  spetrel

  committed
  03140c3

  Copy full SHA for 03140c3

Commits on Feb 28, 2025

• Merge pull request #39 from spetrel/main

  Show description for 2651da3
  unix1986

  authored
  2651da3

  Copy full SHA for 2651da3

• Fix stream decode.

  spetrel

  committed
  e57639b

  Copy full SHA for e57639b

Commits on Feb 27, 2025

• Merge pull request #38 from spetrel/main

  Show description for 4c30a03
  unix1986

  authored
  4c30a03

  Copy full SHA for 4c30a03

• Adapt deepseek_v3

  spetrel

  committed
  971f99d

  Copy full SHA for 971f99d

Commits on Feb 19, 2025

• Merge pull request #37 from yabea/add-reasoner-benchmark

  Show description for 9126bba
  yabea

  authored
  9126bba

  Copy full SHA for 9126bba

• add benchmarking documentation for reasoning models

  

  yangben

  committed
  c36a746

  Copy full SHA for c36a746

Commits on Feb 11, 2025

• fix missing init file in utils (#36)

  Adenialzz

  authored
  a7ec444

  Copy full SHA for a7ec444

• add unified image load util (#35)

  Adenialzz

  authored
  aa0cc61

  Copy full SHA for aa0cc61

Commits on Jan 24, 2025

• [Api] Support logit_bias fix.

  spetrel

  committed
  8f4f714

  Copy full SHA for 8f4f714

• [Api] Support logit_bias

  spetrel

  committed
  630d2bc

  Copy full SHA for 630d2bc

Commits on Jan 23, 2025

• Update README.md

  spetrel

  authored
  3b03c56

  Copy full SHA for 3b03c56

Commits on Jan 22, 2025

• Link flash_attn without IMPORTED_LOCATION.

  spetrel

  committed
  aeb2abd

  Copy full SHA for aeb2abd

• Support LLaMA3 models: Llama-3.1-8B, Llama-3.1-70B, Llama-3.2-1B, Llama-3.2-3B, Llama-3.3-70B

  spetrel

  committed
  754db15

  Copy full SHA for 754db15

Commits on Jan 20, 2025

• Fix deepseek-vl2-tiny config. (#30)

  spetrel

  authored
  30c7643

  Copy full SHA for 30c7643

• Fix test_attention.py (#29)

  spetrel

  authored
  3aab51d

  Copy full SHA for 3aab51d

• Merge pull request #28 from spetrel/main

  Show description for 6b4bd32
  unix1986

  authored
  6b4bd32

  Copy full SHA for 6b4bd32

• Add missed config_util.py

  spetrel

  committed
  8af02f5

  Copy full SHA for 8af02f5

• Simplify test_linear.py (#27)

  spetrel

  authored
  59dc8c2

  Copy full SHA for 59dc8c2

Commits on Jan 16, 2025

• Merge pull request #26 from spetrel/main

  Show description for e36509a
  unix1986

  authored
  e36509a

  Copy full SHA for e36509a

• [Model] Support DeepSeek-VL2 multimodal models: deepseek-vl2-small, deepseek-vl2.

  spetrel

  committed
  34bed64

  Copy full SHA for 34bed64

Pagination

Previous

Next