Commits

Branch selector

 master

User selector

jianyuh

Datepicker

All time

Commit History

Commits on Jul 11, 2020

• [caffe2] Fix the issues when using CUB RadixSort (pytorch#41299)

  Show description for 095886f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  095886f

  Copy full SHA for 095886f

Commits on Jul 10, 2020

• [caffe2] Add the dedup implementation of fused RowWiseAdagrad op on GPUs (pytorch#40282)

  Show description for 62e1693
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  62e1693

  Copy full SHA for 62e1693

Commits on May 13, 2020

• Fix the flooding log issues (pytorch#38356)

  Show description for 2a54533
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  2a54533

  Copy full SHA for 2a54533

Commits on May 5, 2020

• [caffe2] L2 regularization for (RowWise)SparseAdagrad fusion on GPUs (pytorch#37805)

  Show description for 7fa897e
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  7fa897e

  Copy full SHA for 7fa897e

Commits on May 4, 2020

• [TS][easy] Typo Fix (pytorch#37773)

  Show description for fd05deb
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  fd05deb

  Copy full SHA for fd05deb

Commits on Apr 22, 2020

• CUDA implementation of Sparse Adagrad Fusion for GPUs (pytorch#35762)

  Show description for 171476e
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  171476e

  Copy full SHA for 171476e

Commits on Apr 10, 2020

• Move FakeFP16 back to internal to remove dependency on MKL (pytorch#36297)

  Show description for 37c1bd2
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  37c1bd2

  Copy full SHA for 37c1bd2

Commits on Apr 5, 2020

• [pytorch] Fix the extra_repr print message for float16 dynamic quantization (pytorch#36044)

  Show description for 8224398
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  8224398

  Copy full SHA for 8224398

Commits on Feb 25, 2020

• Back out "[pt][quant] RNN debug test" (pytorch#33750)

  Show description for 5ef1c2c
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  5ef1c2c

  Copy full SHA for 5ef1c2c

• [pt][quant] RNN debug test (pytorch#33621)

  Show description for 5b031d9
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  5b031d9

  Copy full SHA for 5b031d9

Commits on Feb 13, 2020

• [pytorch][quant] Add assert for min, max, qmin, qmax for ChooseQuantizationParams (pytorch#32739)

  Show description for 914610d
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  914610d

  Copy full SHA for 914610d

Commits on Jan 30, 2020

• [pytorch][embeddingbag_8bit] Add include_last_offset option to Fused 8bit EmbeddingBag and parallelize the op (pytorch#32683)

  Show description for a840afb
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  a840afb

  Copy full SHA for a840afb

Commits on Jan 29, 2020

• Temporarily disable the test_quantized_rnn test (pytorch#32742)

  Show description for 6f7d5bb
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  6f7d5bb

  Copy full SHA for 6f7d5bb

Commits on Jan 25, 2020

• [pytorch][refactor] Explicitly use auto* for pointers (pytorch#32548)

  Show description for 1218a16
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  1218a16

  Copy full SHA for 1218a16

Commits on Jan 24, 2020

• [pytorch][embeddingbag] Parallelize the EmbeddingBag operator (pytorch#4049)

  Show description for 3ada2e0
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  3ada2e0

  Copy full SHA for 3ada2e0

Commits on Dec 13, 2019

• Use AVX2 to increase frequency for FP16<->FP32 Caffe2 ops (pytorch#31203)

  Show description for 199e1fb
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  199e1fb

  Copy full SHA for 199e1fb

Commits on Dec 12, 2019

• Re-apply "[bert/RoBERTa] Optimize LayerNorm with explicit vectorization using Vec256" (pytorch#31127)

  Show description for 066e3ed
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  066e3ed

  Copy full SHA for 066e3ed

Commits on Dec 11, 2019

• Revert D18293522: Optimize LayerNorm with explicit vectorization using Vec256

  Show description for 4aa30d3
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  4aa30d3

  Copy full SHA for 4aa30d3

• Optimize LayerNorm with explicit vectorization using Vec256 (pytorch#29104)

  Show description for d6d6075
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  d6d6075

  Copy full SHA for d6d6075

Commits on Dec 3, 2019

• Add the explicit per-tensor/per-channel quant info when we print the module (pytorch#30591)

  Show description for 0bebfe2
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  0bebfe2

  Copy full SHA for 0bebfe2

Commits on Nov 21, 2019

• Unit Test for the Legacy Dynamic Quantized Linear operator (pytorch#23139)

  Show description for 38ca355
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  38ca355

  Copy full SHA for 38ca355

• Integrating the int64_t GEMM in FBGEMM into PyTorch Linear op (pytorch#30143)

  Show description for b154a8c
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  b154a8c

  Copy full SHA for b154a8c

Commits on Nov 14, 2019

• Enabling intra-op parallelism for fbgemm_linear_int8_weight_fp32_activation op (pytorch#29532)

  Show description for 4a27d2b
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  4a27d2b

  Copy full SHA for 4a27d2b

Commits on Nov 12, 2019

• Convert conv_prepack to conv2d_prepack and conv_unpack to conv2d_unpack (pytorch#29529)

  Show description for bbff06e
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  bbff06e

  Copy full SHA for bbff06e

Commits on Nov 6, 2019

• Add the intra-op parallelism for equal operator (pytorch#28810)

  Show description for 6a4b51a
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  6a4b51a

  Copy full SHA for 6a4b51a

Commits on Nov 5, 2019

• Enable the intra-op parallelism for layer norm (pytorch#28464)

  Show description for 492764b
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  492764b

  Copy full SHA for 492764b

Commits on Oct 30, 2019

• Update the misleading comments for zero_points and scale in dynamic quant linear module [1/2] (pytorch#28767)

  Show description for b1ea19c
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  b1ea19c

  Copy full SHA for b1ea19c

Commits on Oct 29, 2019

• Avoid the misleading zero_point and scale [2/2] (pytorch#28827)

  Show description for ef5a6b2
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  ef5a6b2

  Copy full SHA for ef5a6b2

• Add the unittest import for test_fake_quant.py (pytorch#28815)

  Show description for 6f90567
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  6f90567

  Copy full SHA for 6f90567

Commits on Oct 28, 2019

• Add the warning message for API with linear modules (pytorch#28766)

  Show description for 9e64c54
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  9e64c54

  Copy full SHA for 9e64c54

• Temporarily disable test_numerical_consistency_per_channel due to failure (pytorch#28807)

  Show description for 02d3184
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  02d3184

  Copy full SHA for 02d3184

• Enabling intra-op parallelism for dynamic quantized Linear operator (pytorch#28477)

  Show description for 052046b
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  052046b

  Copy full SHA for 052046b

Commits on Oct 1, 2019

• Uninitialize the accumulation buffer to save some overhead (pytorch#27005)

  Show description for 6a09676
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  6a09676

  Copy full SHA for 6a09676

Commits on Sep 23, 2019

• Fix the flaky test_qlinear test caused by hypothesis deadline (pytorch#26663)

  Show description for cbdbdd3
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  cbdbdd3

  Copy full SHA for cbdbdd3

Commits on Sep 20, 2019

• Add the FP16 weight support for LSTM in dynamic_quantize (pytorch#25975)

  Show description for f433ee1
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  f433ee1

  Copy full SHA for f433ee1

Pagination

Previous

Next