Commits

Branch selector

 3bc185b

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 25, 2022

• OSS fbgemm_gpu: Try to automatically find correct cuda installation in setup.py (pytorch#879)

  Show description for 3bc185b
  

  Stephan Uphoff

  authored and

  facebook-github-bot

  committed
  3bc185b

Commits on Jan 24, 2022

• use self.Assert in split_table_batched_embeddings_test (pytorch#880)

  Show description for ba72edd
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  ba72edd

Commits on Jan 21, 2022

• Transposed Direct Convolution for kh=2 kw=6

  Show description for 81719d6
  

  jiyuanzFB

  authored and

  facebook-github-bot

  committed
  81719d6

• Add support for different types for segment value vs. length. (pytorch#878)

  Show description for 0406bf6
  

  jasonjk-park

  authored and

  facebook-github-bot

  committed
  0406bf6

• Add FBGEMM_GPU unit tests to GHA CPU-only build. (pytorch#875)

  Show description for 1b6412a
  

  Rick Weyrauch

  authored and

  facebook-github-bot

  committed
  1b6412a

Commits on Jan 20, 2022

• fbgemm_gpu: Remove unused file (pytorch#877)

  Show description for bef5415
  

  Stephan Uphoff

  authored and

  facebook-github-bot

  committed
  bef5415

• reduce memory used for test_stacked_jagged_2d_to_dense take 2 (pytorch#876)

  Show description for 17d6a15
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  17d6a15

• Back out "add bfloat16 gradient option to split_table_batched_embeddings_benchmark (2/2)"

  Show description for 8275eb6
  

  Sudarshan Raghunathan

  authored and

  facebook-github-bot

  committed
  8275eb6

• enable -Wall -Wextra -Werror (pytorch#868)

  Show description for b047231
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  b047231

• Support other integral types for segment values in HBC by feature (pytorch#874)

  Show description for 9e8820b
  

  jasonjk-park

  authored and

  facebook-github-bot

  committed
  9e8820b

Commits on Jan 19, 2022

• add bfloat16 gradient option to split_table_batched_embeddings_benchmark (2/2)

  Show description for 48515f9
  

  shintaro-iwasaki

  authored and

  facebook-github-bot

  committed
  48515f9

• Update OSS README for FBGEMM_GPU (pytorch#869)

  Show description for 41ebf9b
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  41ebf9b

• Save OSS build time by multi-threading; Fix OSS build failure (pytorch#865)

  Show description for 278143a
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  278143a

Commits on Jan 18, 2022

• uvm bench update (pytorch#873)

  Show description for 36c8343
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  36c8343

• Update GHA CI build to install fbgemm_gpu as root. (pytorch#872)

  Show description for 9576b0c
  

  Rick Weyrauch

  authored and

  facebook-github-bot

  committed
  9576b0c

• Implement cpu-only build support into the cmake build. (pytorch#870)

  Show description for 0fd8dbd
  

  Rick Weyrauch

  authored and

  facebook-github-bot

  committed
  0fd8dbd

• Integrate fbgemm_gpu build with fbgemm Github build. (pytorch#862)

  Show description for 2115e01
  

  Rick Weyrauch

  authored and

  facebook-github-bot

  committed
  2115e01

• remove extra ; after marco (pytorch#866)

  Show description for 80e3f16
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  80e3f16

• actually clamp denorm in RoundToFloat16 when clamp_denorms is set (pytorch#871)

  Show description for c3ad33c
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  c3ad33c

• reduce memory used for test_stacked_jagged_2d_to_dense (pytorch#867)

  Show description for 9843c86
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  9843c86

Commits on Jan 14, 2022

• Skip long-running (large-data) tests in CI environment. (pytorch#864)

  Show description for f6fd973
  

  Rick Weyrauch

  authored and

  facebook-github-bot

  committed
  f6fd973

• Remove dead includes in deeplearning/fbgemm/fbgemm_gpu

  Show description for 1d09a3d
  

  Dead Code Bot

  authored and

  facebook-github-bot

  committed
  1d09a3d

Commits on Jan 13, 2022

• move ATen/CUDAGeneratorImpl.h to ATen/cuda (#71224)

  Show description for eb17d45
  

  shintaro-iwasaki

  authored and

  facebook-github-bot

  committed
  eb17d45

• conv code refactoring (pytorch#856)

  Show description for 3fac636
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  3fac636

• Disable clang inline asm workaround for clang-14 and later (pytorch#861)

  Show description for 135412d
  

  MatzeB

  authored and

  facebook-github-bot

  committed
  135412d

Commits on Jan 12, 2022

• Fix OSS license (pytorch#858)

  Show description for 9d7c48a
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  9d7c48a

• optimize ESUHM compute by fusing embedding lookups and padding

  Show description for 5b945ac
  

  xing-liu

  authored and

  facebook-github-bot

  committed
  5b945ac

• Continue moving fbgemm_gpu operators to the preferred fbgemm namespace. (pytorch#859)

  Show description for 2f6c894
  

  Rick Weyrauch

  authored and

  facebook-github-bot

  committed
  2f6c894

Commits on Jan 10, 2022

• Update OSS license (pytorch#855)

  Show description for 22d234e
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  22d234e

Commits on Jan 9, 2022

• mixed dim int8 dequantize benchmark (pytorch#854)

  Show description for e3b52c7
  

  Mengchi Zhang

  authored and

  facebook-github-bot

  committed
  e3b52c7

Commits on Jan 8, 2022

• int8 dequantize kernel for batched mixed dimension pooled embedding output (pytorch#853)

  Show description for 3c5b494
  

  Mengchi Zhang

  authored and

  facebook-github-bot

  committed
  3c5b494

• Tiny cleanups (pytorch#852)

  Show description for a0347b2
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  a0347b2

• use fbgemm cpu kernel in forward quantized cpu emb (pytorch#849)

  Show description for 0e5737d
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  0e5737d

• scale_bias_last option to support TBE layout (pytorch#848)

  Show description for 747fc4a
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  747fc4a

Commits on Jan 7, 2022

• Support INT8 -> Half conversion to reduce the dequantize overhead (pytorch#851)

  Show description for fa20cb3
  

  Rick Weyrauch

  authored and

  facebook-github-bot

  committed
  fa20cb3

Pagination

PreviousNext