Commits

Branch selector

 main

User selector

jianyuh

Datepicker

All time

Commit History

Commits on Sep 6, 2024

• EP support (pytorch#3071)

  Show description for 0dc097f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  0dc097f

Commits on Aug 20, 2024

• move mqa code (pytorch#3011)

  Show description for c9ad39e
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  c9ad39e

Commits on Aug 19, 2024

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

  Show description for f6a282e
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  f6a282e

Commits on Aug 16, 2024

• refactor kv cache ops (pytorch#3001)

  Show description for 4e207dd
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  4e207dd

Commits on Aug 15, 2024

• Add kv cache related ops (pytorch#2968)

  Show description for 7105d5f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  7105d5f

Commits on Aug 12, 2024

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

  Show description for 6a3ff16
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  6a3ff16

Commits on Aug 5, 2024

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

  Show description for 3dd81d5
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  3dd81d5

Commits on Aug 2, 2024

• Add README for fbgemm genAI op (pytorch#2931)

  Show description for 3a5bb1d
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  3a5bb1d

Commits on Jul 25, 2024

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

  Show description for 29c8f33
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  29c8f33

Commits on Jul 24, 2024

• Fix unit test failure on test_quantize_fp8_matmul (pytorch#2869)

  Show description for 9a8d114
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  9a8d114

Commits on Jul 23, 2024

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

  Show description for d313ac7
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  d313ac7

Commits on Jul 19, 2024

• Fix bf16i4bf16 unit test failure (pytorch#2864)

  Show description for 521c3ad
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  521c3ad

Commits on Jul 16, 2024

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

  Show description for 903e928
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  903e928

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

  Show description for da410c0
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  da410c0

Commits on Jul 12, 2024

• remove unused exception (pytorch#2832)

  Show description for 3d6e0c7
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  3d6e0c7

Commits on Jul 4, 2024

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

  Show description for f3b33f9
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  f3b33f9

Commits on Jun 18, 2024

• update shapes

  Show description for 39b2108
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  39b2108

Commits on Jun 17, 2024

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

  Show description for 74fc8d3
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  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

Commits on Jun 3, 2024

• Tiny typo fix (pytorch#2662)

  Show description for 293a30b
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  293a30b

Commits on May 23, 2024

• Add car comm ops (pytorch#2606)

  Show description for 4c3af94
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  4c3af94

Commits on May 22, 2024

• Reuse cuda_calc_xblock_count and remove duplicated copy (pytorch#2607)

  Show description for 8f19539
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  8f19539

Commits on May 19, 2024

• Fix gen_ai only build (pytorch#2603)

  Show description for 37c283c
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  37c283c

Commits on May 15, 2024

• Add GenAI build only option (pytorch#2584)

  Show description for 50d6808
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  50d6808

Commits on May 6, 2024

• use verbose mode for CMake (pytorch#2563)

  Show description for ca71f4a
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  ca71f4a

Commits on May 5, 2024

• Clean up llama csrc code base

  Show description for bc4e9c3
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  bc4e9c3

Commits on May 1, 2024

• Dedup GQA splitk kernel

  Show description for 0aecd17
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  0aecd17

Commits on Apr 30, 2024

• Refactor fbgemm / llama csrc code base

  Show description for 0da2f0c
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  0da2f0c

• Update contbuild/citadel configuration to use h100a in CI

  Show description for ee1c4a1
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  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

• Enhance AMD support

  Show description for 53e7da5
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  53e7da5

• Host commonly used quantize (fp8 tensor/row/col wise quant) op in fbgemm

  Show description for 39b655a
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  39b655a

Commits on Feb 20, 2024

• Use gpu_library_selector for permute_pooled_embedding_ops_gpu (pytorch#2340)

  Show description for 49cca04
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  49cca04

Commits on Jan 1, 2024

• Address half->int build error (pytorch#2240)

  Show description for 905ff66
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  905ff66

Commits on Dec 28, 2023

• Fix bf16 support issues (pytorch#2238)

  Show description for 9cd944a
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  9cd944a

Pagination

PreviousNext