Commits

Branch selector

 master

User selector

yufenglee

Datepicker

All time

Commit History

Commits on Apr 23, 2020

• Fix a bug in skiplayernorm fusion pattern 2 (#3660)

  Show description for c0e817f
  yufenglee

  authored
  c0e817f

  Copy full SHA for c0e817f

Commits on Apr 17, 2020

• Make De/QuantizeLinear support half (#3531)

  Show description for f822a54
  yufenglee

  authored
  f822a54

  Copy full SHA for f822a54

Commits on Apr 14, 2020

• Handle the case that initializers are in graph input (#3449)

  Show description for baa86f1
  yufenglee

  authored
  baa86f1

  Copy full SHA for baa86f1

Commits on Apr 9, 2020

• Revert "Use IMMA for int8 matmul to leverage Turing Tensor Core (#3413)" (#3472)

  Show description for a443b1b
  yufenglee

  authored
  a443b1b

  Copy full SHA for a443b1b

Commits on Apr 7, 2020

• Use IMMA for int8 matmul to leverage Turing Tensor Core (#3413)

  Show description for 4d71958
  yufenglee

  authored
  4d71958

  Copy full SHA for 4d71958

Commits on Mar 31, 2020

• fix bugs in quantization and calibration tools (#3329)

  Show description for af61827
  yufenglee

  authored
  af61827

  Copy full SHA for af61827

Commits on Mar 20, 2020

• fix quantize_bias (#3270)

  yufenglee

  authored
  a69d859

  Copy full SHA for a69d859

Commits on Mar 13, 2020

• fix the missing return in _get_quantize_input_nodes and format code with yapf (#3199)

  Show description for c69194e
  yufenglee

  authored
  c69194e

  Copy full SHA for c69194e

Commits on Mar 10, 2020

• Move zero point inputs of MatmulInteger to CPU memory (#3159)

  yufenglee

  authored
  3de1fc0

  Copy full SHA for 3de1fc0

• Make quantization parameters as constant weigth instead of overrideable (#3160)

  yufenglee

  authored
  876d0c5

  Copy full SHA for 876d0c5

Commits on Mar 5, 2020

• Support u8u8 in quantization tool (#3140)

  yufenglee

  authored
  1d2b811

  Copy full SHA for 1d2b811

Commits on Mar 4, 2020

• Implement QuantizeLinear and DequantizeLinear (#3098)

  Show description for fbb658e
  yufenglee

  authored
  fbb658e

  Copy full SHA for fbb658e

• Implement MatmulInteger on GPU (#3070)

  Show description for 84ad4ed
  yufenglee

  authored
  84ad4ed

  Copy full SHA for 84ad4ed

Commits on Feb 27, 2020

• Bump up version number to 1.2 (#3097)

  yufenglee

  authored
  ca2ed17

  Copy full SHA for ca2ed17

Commits on Feb 26, 2020

• Disable test_zfnet512 and test_bvlc_reference_caffenet for x86 in C# tests (#3094)

  yufenglee

  authored
  f1ba531

  Copy full SHA for f1ba531

Commits on Feb 22, 2020

• Logging in framework.cc should use the session logger (#3059)

  yufenglee

  authored
  64feee1

  Copy full SHA for 64feee1

Commits on Jan 31, 2020

• Add int64_t support for split (#2944)

  Show description for 1d79926
  yufenglee

  authored
  1d79926

  Copy full SHA for 1d79926

Commits on Jan 30, 2020

• keep output of bert model as float32 when converting to float16 (#2913)

  yufenglee

  authored
  f337598

  Copy full SHA for f337598

Commits on Jan 24, 2020

• Only fuse when output count of add is 1 (#2884)

  Show description for cd87672
  yufenglee

  authored
  cd87672

  Copy full SHA for cd87672

Commits on Jan 18, 2020

• Add float16 support back in the bert fusion script (#2870)

  Show description for 25d7ad1
  yufenglee

  authored
  25d7ad1

  Copy full SHA for 25d7ad1

Commits on Jan 17, 2020

• Bert fusion script for Tensorflow squad (#2858)

  yufenglee

  authored
  95f3eb6

  Copy full SHA for 95f3eb6

Commits on Jan 9, 2020

• Initialize max of softmax with lowest of float (#2786)

  yufenglee

  authored
  71b5165

  Copy full SHA for 71b5165

Commits on Jan 6, 2020

• Implement a more stable softmax (#2715)

  Show description for 72bdfc8
  yufenglee

  authored
  72bdfc8

  Copy full SHA for 72bdfc8

Commits on Dec 20, 2019

• fix the CUDNN_BN_MIN_EPSILON difference issue between cudnn7.3 and cudnn7.6 (#2681)

  yufenglee

  authored
  72286d9

  Copy full SHA for 72286d9

Commits on Dec 12, 2019

• fix float16 comparison in initializer (#2629)

  

  yufenglee

  authored and

  tianleiwu

  committed
  d6f33dc

  Copy full SHA for d6f33dc

Commits on Dec 6, 2019

• Add more check on SkipLayerNorm and BiasGelu fusion (#2574)

  yufenglee

  authored
  5575766

  Copy full SHA for 5575766

• make layernorm fusion to support opset 11 (#2545)

  yufenglee

  authored
  34beafc

  Copy full SHA for 34beafc

Commits on Nov 28, 2019

• Implement AddGelu and SkipLayerNorm (#2487)

  Show description for 005305b
  yufenglee

  authored
  005305b

  Copy full SHA for 005305b

Commits on Nov 22, 2019

• Fuse SkipLayerNorm with Bias (#2453)

  Show description for b3bfc09
  yufenglee

  authored
  b3bfc09

  Copy full SHA for b3bfc09

• Parallel Gelu with ParallelFor (#2399)

  Show description for d49cbf6
  yufenglee

  authored
  d49cbf6

  Copy full SHA for d49cbf6

Commits on Nov 13, 2019

• Fuse Add + Gelu (#2360)

  Show description for 8ed2928
  yufenglee

  authored
  8ed2928

  Copy full SHA for 8ed2928

Commits on Nov 7, 2019

• Make elementwise op run 4 items per thread (#2335)

  Show description for 6651d2f
  yufenglee

  authored
  6651d2f

  Copy full SHA for 6651d2f

Commits on Nov 5, 2019

• Support int32_t for Reduction (#2317)

  yufenglee

  authored
  035913d

  Copy full SHA for 035913d

Commits on Oct 15, 2019

• use cublasHgemm for Volta GPU (#2074)

  Show description for 2536553
  yufenglee

  authored
  2536553

  Copy full SHA for 2536553

Commits on Oct 14, 2019

• use legacy stream mode (#2076)

  Show description for 8c5db7f
  yufenglee

  authored
  8c5db7f

  Copy full SHA for 8c5db7f

Pagination

Previous

Next