Commits

Branch selector

 main

User selector

jianyuh

Datepicker

All time

Commit History

Commits on Jun 16, 2025

• Add FP32 support for routing_score dtype (#4352)

  Show description for 8a01350
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  8a01350

  Copy full SHA for 8a01350

Commits on May 16, 2025

• Add Llama4 shapes in quantize_bench (#4129)

  Show description for 1e9425f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  1e9425f

  Copy full SHA for 1e9425f

Commits on Apr 21, 2025

• Update fbgemm gen_ai README (#3997)

  Show description for c0052af
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  c0052af

  Copy full SHA for c0052af

Commits on Mar 19, 2025

• Replace rms_norm as norm (#3841)

  Show description for 89c2d0d
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  89c2d0d

  Copy full SHA for 89c2d0d

Commits on Sep 6, 2024

• EP support (#3071)

  Show description for 0dc097f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  0dc097f

  Copy full SHA for 0dc097f

Commits on Aug 20, 2024

• move mqa code (#3011)

  Show description for c9ad39e
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  c9ad39e

  Copy full SHA for c9ad39e

Commits on Aug 19, 2024

• Fix the dev non-san fbgemm op loading issue (#3012)

  Show description for f6a282e
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  f6a282e

  Copy full SHA for f6a282e

Commits on Aug 16, 2024

• refactor kv cache ops (#3001)

  Show description for 4e207dd
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  4e207dd

  Copy full SHA for 4e207dd

Commits on Aug 15, 2024

• Add kv cache related ops (#2968)

  Show description for 7105d5f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  7105d5f

  Copy full SHA for 7105d5f

Commits on Aug 12, 2024

• include fb only ops in private fbgemm build (#2972)

  Show description for 6a3ff16
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  6a3ff16

  Copy full SHA for 6a3ff16

Commits on Aug 5, 2024

• Back out "Optimize the cuda rowwise quantize kernel" (#2934)

  Show description for 3dd81d5
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  3dd81d5

  Copy full SHA for 3dd81d5

Commits on Aug 2, 2024

• Add README for fbgemm genAI op (#2931)

  Show description for 3a5bb1d
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  3a5bb1d

  Copy full SHA for 3a5bb1d

Commits on Jul 25, 2024

• Fix customized all reduce device mismatch issue on RoCM 6.2.0 (#2876)

  Show description for 29c8f33
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  29c8f33

  Copy full SHA for 29c8f33

Commits on Jul 24, 2024

• Fix unit test failure on test_quantize_fp8_matmul (#2869)

  Show description for 9a8d114
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  9a8d114

  Copy full SHA for 9a8d114

Commits on Jul 23, 2024

• Fix multi gpu customized all reduce test for AMD (#2875)

  Show description for d313ac7
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  d313ac7

  Copy full SHA for d313ac7

Commits on Jul 19, 2024

• Fix bf16i4bf16 unit test failure (#2864)

  Show description for 521c3ad
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  521c3ad

  Copy full SHA for 521c3ad

Commits on Jul 16, 2024

• Add 3D+ input support for fp8 rowwise GEMM (#2845)

  Show description for 903e928
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  903e928

  Copy full SHA for 903e928

• Hipify the fp8 quantize related cuda functions (#2834)

  Show description for da410c0
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  da410c0

  Copy full SHA for da410c0

Commits on Jul 12, 2024

• remove unused exception (#2832)

  Show description for 3d6e0c7
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  3d6e0c7

  Copy full SHA for 3d6e0c7

Commits on Jul 4, 2024

• Fix the subnormal subnormal adjustment for FP8 stochastic rounding (#2798)

  Show description for f3b33f9
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  f3b33f9

  Copy full SHA for f3b33f9

Commits on Jun 18, 2024

• update shapes

  Show description for 39b2108
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  39b2108

  Copy full SHA for 39b2108

Commits on Jun 17, 2024

• Support stochastic rounding for FP32/BF16 -> FP8 conversion (#2677)

  Show description for 74fc8d3
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  74fc8d3

  Copy full SHA for 74fc8d3

Commits on Jun 10, 2024

• Revert D58147817: Multisect successfully blamed "D58147817: [FBGEMM][PR] [fbgemm_gpu] Enable NCCL code" for one test failure

  Show description for d9d70ab
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  d9d70ab

  Copy full SHA for d9d70ab

Commits on Jun 3, 2024

• Tiny typo fix (#2662)

  Show description for 293a30b
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  293a30b

  Copy full SHA for 293a30b

Commits on May 23, 2024

• Add car comm ops (#2606)

  Show description for 4c3af94
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  4c3af94

  Copy full SHA for 4c3af94

Commits on May 22, 2024

• Reuse cuda_calc_xblock_count and remove duplicated copy (#2607)

  Show description for 8f19539
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  8f19539

  Copy full SHA for 8f19539

Commits on May 19, 2024

• Fix gen_ai only build (#2603)

  Show description for 37c283c
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  37c283c

  Copy full SHA for 37c283c

Commits on May 15, 2024

• Add GenAI build only option (#2584)

  Show description for 50d6808
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  50d6808

  Copy full SHA for 50d6808

Commits on May 6, 2024

• use verbose mode for CMake (#2563)

  Show description for ca71f4a
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  ca71f4a

  Copy full SHA for ca71f4a

Commits on May 5, 2024

• Clean up llama csrc code base

  Show description for bc4e9c3
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  bc4e9c3

  Copy full SHA for bc4e9c3

Commits on May 1, 2024

• Dedup GQA splitk kernel

  Show description for 0aecd17
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  0aecd17

  Copy full SHA for 0aecd17

Commits on Apr 30, 2024

• Refactor fbgemm / llama csrc code base

  Show description for 0da2f0c
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  0da2f0c

  Copy full SHA for 0da2f0c

• Update contbuild/citadel configuration to use h100a in CI

  Show description for ee1c4a1
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  ee1c4a1

  Copy full SHA for ee1c4a1

Commits on Apr 29, 2024

• Change model transform fp8 linear op to fbgemm quantize ops

  Show description for dff5bc2
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  dff5bc2

  Copy full SHA for dff5bc2

• Enhance AMD support

  Show description for 53e7da5
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  53e7da5

  Copy full SHA for 53e7da5

Pagination

Previous

Next