Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 28, 2025

• Merge pull request #88 from ROCm/IFU-20250128

  Show description for b97e51c
  liligwu

  authored
  b97e51c

  Copy full SHA for b97e51c

• Merge remote-tracking branch 'upstream/main' into IFU-20250128

  liligwu

  committed
  e12fc5e

  Copy full SHA for e12fc5e

• Move embedding_rocksdb_wrapper to its own header. (pytorch#3622)

  Show description for bab9b62
  

  pradeepfn

  authored and

  facebook-github-bot

  committed
  bab9b62

  Copy full SHA for bab9b62

• fix pre_iter fp32 inaccuracy issue (pytorch#3623)

  Show description for f7d5dae
  

  Wang Zhou

  authored and

  facebook-github-bot

  committed
  f7d5dae

  Copy full SHA for f7d5dae

Commits on Jan 27, 2025

• CK MoE: cherry-pick pytorch#1808 (pytorch#3609)

  Show description for f92c108
  

  coconutruben

  authored and

  facebook-github-bot

  committed
  f92c108

  Copy full SHA for f92c108

• directly pass update_util as int flag without syncing iter (pytorch#3602)

  Show description for 7e0d930
  

  Fei Yu

  authored and

  facebook-github-bot

  committed
  7e0d930

  Copy full SHA for 7e0d930

• Improve handling for FP8 grouped gemm without zero_start_index_M (pytorch#3615)

  Show description for c8da60b
  

  jwfromm

  authored and

  facebook-github-bot

  committed
  c8da60b

  Copy full SHA for c8da60b

Commits on Jan 25, 2025

• basic tbe input dump framework (pytorch#3593)

  Show description for e7c08e3
  

  Sihui Han

  authored and

  facebook-github-bot

  committed
  e7c08e3

  Copy full SHA for e7c08e3

• add AMD specific includes in cuda_prelude.h (pytorch#3614)

  Show description for f9650f9
  

  Bangsheng Tang

  authored and

  facebook-github-bot

  committed
  f9650f9

  Copy full SHA for f9650f9

Commits on Jan 24, 2025

• fix llm shapes in quantize bench and add ldm shapes (pytorch#3611)

  Show description for 1aff241
  

  mxz297

  authored and

  facebook-github-bot

  committed
  1aff241

  Copy full SHA for 1aff241

• amd fp8 rowwise gemm prefill shape tuning (pytorch#3607)

  Show description for 74490d6
  

  mxz297

  authored and

  facebook-github-bot

  committed
  74490d6

  Copy full SHA for 74490d6

• Add support for int32_t indices in TBE training (2K/N) (pytorch#3583)

  Show description for 5f3adca
  

  q10

  authored and

  facebook-github-bot

  committed
  5f3adca

  Copy full SHA for 5f3adca

Commits on Jan 23, 2025

• Enable fast FP8 GEMM for memory bound (resubmit) (pytorch#3608)

  Show description for 5754ce7
  

  jiawenliu64

  authored and

  facebook-github-bot

  committed
  5754ce7

  Copy full SHA for 5754ce7

• add a new constructor to PackedGemmMatrixB (pytorch#3598)

  Show description for 31d41dc
  

  helloguo

  authored and

  facebook-github-bot

  committed
  31d41dc

  Copy full SHA for 31d41dc

• Pre-convert indices/offsets in TBE bench (pytorch#3595)

  Show description for 6a1cadb
  

  sryap

  authored and

  facebook-github-bot

  committed
  6a1cadb

  Copy full SHA for 6a1cadb

• Make zero_start_index_M optional for dynamic FP8 grouped gemm on AMD (pytorch#3604)

  Show description for 121e8ba
  

  jiawenliu64

  authored and

  facebook-github-bot

  committed
  121e8ba

  Copy full SHA for 121e8ba

• Add support for int32_t indices in TBE training (2H/N) (pytorch#3539)

  Show description for fde11cd
  

  q10

  authored and

  facebook-github-bot

  committed
  fde11cd

  Copy full SHA for fde11cd

Commits on Jan 22, 2025

• Fix bug in ROCm optimized forward pass (pytorch#3599)

  Show description for 929aa82
  

  avbokovoy

  authored and

  facebook-github-bot

  committed
  929aa82

  Copy full SHA for 929aa82

• Fix type to assert group size (pytorch#3601)

  Show description for e5439b8
  

  Robert Luo

  authored and

  facebook-github-bot

  committed
  e5439b8

  Copy full SHA for e5439b8

• Updates and fixes to tensor_accessor.h (2I/N) (pytorch#3571)

  Show description for 4df8a12
  

  q10

  authored and

  facebook-github-bot

  committed
  4df8a12

  Copy full SHA for 4df8a12

• Fix CMakeLists.txt for experimental/gemm (pytorch#3592)

  Show description for 559945b
  

  q10

  authored and

  facebook-github-bot

  committed
  559945b

  Copy full SHA for 559945b

• Fix autodeps for torch/custom_class.h and use it in kv_tensor_wrapper_cpu (pytorch#3600)

  Show description for 6dbc368
  

  r-barnes

  authored and

  facebook-github-bot

  committed
  6dbc368

  Copy full SHA for 6dbc368

• Return if no data to allreduce (pytorch#3586)

  Show description for 6a684cb
  

  xw285cornell

  authored and

  facebook-github-bot

  committed
  6a684cb

  Copy full SHA for 6a684cb

• Add support for int32_t indices in TBE training (2I/N) (pytorch#3556)

  Show description for b858408
  

  q10

  authored and

  facebook-github-bot

  committed
  b858408

  Copy full SHA for b858408

• Add support for int32_t indices in TBE training (2F/N) (pytorch#3376)

  Show description for 79a75bf
  

  q10

  authored and

  facebook-github-bot

  committed
  79a75bf

  Copy full SHA for 79a75bf

• Allow reusing input data in TBE benchmark (pytorch#3594)

  Show description for 51d2be6
  

  sryap

  authored and

  facebook-github-bot

  committed
  51d2be6

  Copy full SHA for 51d2be6

Commits on Jan 21, 2025

• fix data type of block_bucketize_pos in block_bucketize_sparse_features (pytorch#3589)

  Show description for 5852de7
  

  TroyGarden

  authored and

  facebook-github-bot

  committed
  5852de7

  Copy full SHA for 5852de7

• Back out "Optimzed backward pass for ROCm devices (pt 2)" (pytorch#3587)

  Show description for 8b02cbe
  

  q10

  authored and

  facebook-github-bot

  committed
  8b02cbe

  Copy full SHA for 8b02cbe

• Revert D65620886 (pytorch#3582)

  Show description for 87db593
  

  q10

  authored and

  facebook-github-bot

  committed
  87db593

  Copy full SHA for 87db593

Commits on Jan 20, 2025

• Merge pull request #86 from ROCm/IFU-20250120

  Show description for d5b7155
  liligwu

  authored
  d5b7155

  Copy full SHA for d5b7155

• Merge remote-tracking branch 'upstream/main' into IFU-20250120

  liligwu

  committed
  ad6a1fb

  Copy full SHA for ad6a1fb

Commits on Jan 17, 2025

• Back out "Enable fast FP8 GEMM for memory bound" (pytorch#3588)

  Show description for 2d025dc
  

  jiawenliu64

  authored and

  facebook-github-bot

  committed
  2d025dc

  Copy full SHA for 2d025dc

• Profile with kineto and warmup for more accurate benchmarking (pytorch#3585)

  Show description for 379db5f
  

  amirakb89

  authored and

  facebook-github-bot

  committed
  379db5f

  Copy full SHA for 379db5f

• Fix specailization issue in keyed_jagged_index_select_dim1_forward_cuda (pytorch#3578)

  Show description for 21d1260
  

  Microve

  authored and

  facebook-github-bot

  committed
  21d1260

  Copy full SHA for 21d1260

Commits on Jan 16, 2025

• Upgrade GitHub actions (pytorch#3581)

  Show description for 3e0db25
  

  q10

  authored and

  facebook-github-bot

  committed
  3e0db25

  Copy full SHA for 3e0db25

Pagination

Previous

Next