Commits

Branch selector

 develop

User selector

GuoxiaWang

Datepicker

All time

Commit History

Commits on Mar 28, 2024

• support flash attention with sparse mask (PaddlePaddle#62029)

  Show description for e05764a
  

  GuoxiaWang

  and

  sunzhongkai588

  authored
  e05764a

  Copy full SHA for e05764a

Commits on Apr 13, 2023

• add uint16 for bfloat16 dtype check in layer_norm under static mode (PaddlePaddle#52845)

  GuoxiaWang

  authored
  5a86427

  Copy full SHA for 5a86427

Commits on Apr 12, 2023

• [AMP OP&Test] support bf16 for batch norm (PaddlePaddle#52407)

  Show description for 523f8a2
  GuoxiaWang

  authored
  523f8a2

  Copy full SHA for 523f8a2

Commits on Mar 30, 2023

• support python object input data broadcast for model parallel (PaddlePaddle#51765)

  Show description for 8baf33a
  GuoxiaWang

  authored
  8baf33a

  Copy full SHA for 8baf33a

Commits on Mar 29, 2023

• [AMP OP&Test] update bf16 atol=0.1 to atol=0.02 (PaddlePaddle#52121)

  GuoxiaWang

  authored
  fa07857

  Copy full SHA for fa07857

Commits on Mar 15, 2023

• add inplace sigmoid_ and multiply_ (PaddlePaddle#50267)

  GuoxiaWang

  authored
  b3caa23

  Copy full SHA for b3caa23

Commits on Sep 15, 2022

• remove tmp fp32 var for gaussian_random (PaddlePaddle#46033)

  GuoxiaWang

  authored
  3671d11

  Copy full SHA for 3671d11

Commits on Jun 15, 2022

• modify index dtype from int to int64_t of concat_and_split_functor (PaddlePaddle#43479)

  GuoxiaWang

  authored
  81abaaf

  Copy full SHA for 81abaaf

Commits on Jun 7, 2022

• add bf16 dtype for flatten kernel (PaddlePaddle#43264)

  GuoxiaWang

  authored
  0fdb3ce

  Copy full SHA for 0fdb3ce

• fix the unittest bug of none grad of margin_cross_entropy when FLAGS_retain_grad_for_all_tensor change default setting (PaddlePaddle#43241)

  GuoxiaWang

  authored
  e74f287

  Copy full SHA for e74f287

Commits on Jun 2, 2022

• fix the bug of margin cross entropy loss for eager mode (PaddlePaddle#43161)

  GuoxiaWang

  authored
  67163fb

  Copy full SHA for 67163fb

Commits on Jun 1, 2022

• support nccl api for bfloat16, required >= cudnn 10.1, nccl >= 2.10.3 (PaddlePaddle#43147)

  GuoxiaWang

  authored
  67b9b51

  Copy full SHA for 67b9b51

• fix the bug of adamw which set the attribute in param group not working (PaddlePaddle#43013)

  Show description for 77bae9a
  GuoxiaWang

  authored
  77bae9a

  Copy full SHA for 77bae9a

Commits on Mar 24, 2022

• support dp for class_center_sample and margin_cross_entropy (PaddlePaddle#39852)

  GuoxiaWang

  authored
  bff9e28

  Copy full SHA for bff9e28

Commits on Mar 1, 2022

• add MasterParam and MasterParamOut for sparse_momentum op (PaddlePaddle#39969)

  GuoxiaWang

  authored
  9de7989

  Copy full SHA for 9de7989

Commits on Jan 10, 2022

• remove fp32 tmp tensor and cast op for initializer.Normal and initializer.Constant (PaddlePaddle#38818)

  GuoxiaWang

  authored
  2238a53

  Copy full SHA for 2238a53

• fix cuda seed bug of class_center_sample traning on multi gpu (PaddlePaddle#38815)

  GuoxiaWang

  authored
  04f73d8

  Copy full SHA for 04f73d8

Commits on Dec 28, 2021

• fix adamw epsilon in cuda kernel (PaddlePaddle#37746)

  GuoxiaWang

  authored
  6f1bb3d

  Copy full SHA for 6f1bb3d

Commits on Dec 27, 2021

• gelu using normcdf for cudnn (PaddlePaddle#38450)

  GuoxiaWang

  authored
  3702248

  Copy full SHA for 3702248

Commits on Dec 22, 2021

• fix prelu weight shape for NHWC of static mode (PaddlePaddle#38310)

  GuoxiaWang

  authored
  0a79499

  Copy full SHA for 0a79499

Commits on Dec 21, 2021

• fix recompute no grad warning (PaddlePaddle#38293)

  GuoxiaWang

  authored
  2005b98

  Copy full SHA for 2005b98

Commits on Dec 18, 2021

• fix seed for class_center_sample using paddle.seed (PaddlePaddle#38248)

  GuoxiaWang

  authored
  59be8e0

  Copy full SHA for 59be8e0

Commits on Nov 30, 2021

• support data_format='NHWC' for prelu channel mode (PaddlePaddle#37019)

  Show description for 3f2a665
  GuoxiaWang

  authored
  3f2a665

  Copy full SHA for 3f2a665

Commits on Nov 8, 2021

• fix bug for fp32 batchnorm_op when using nhwc data_layout (PaddlePaddle#37020)

  GuoxiaWang

  authored
  0666b85

  Copy full SHA for 0666b85

Commits on Oct 13, 2021

• fix BatchNorm for fp16 (PaddlePaddle#36376)

  Show description for 8fd1b6a
  GuoxiaWang

  authored
  8fd1b6a

  Copy full SHA for 8fd1b6a

• support auto parallel data shard (PaddlePaddle#36055)

  GuoxiaWang

  authored
  85bb1a8

  Copy full SHA for 85bb1a8

Commits on Sep 28, 2021

• fix bug of reduce_sum when src_dtype != dst_dtype and reduce_num == 1 (PaddlePaddle#36123)

  GuoxiaWang

  authored
  d5268a6

  Copy full SHA for d5268a6

Commits on Sep 21, 2021

• support fp16 (PaddlePaddle#35888)

  GuoxiaWang

  authored
  087c23a

  Copy full SHA for 087c23a

Commits on Sep 18, 2021

• fix bug of module 'paddle' has no attribute 'distributed' for python3.6 (PaddlePaddle#35848)

  Show description for d4cd259
  GuoxiaWang

  authored
  d4cd259

  Copy full SHA for d4cd259

Commits on Sep 17, 2021

• fix unittest (PaddlePaddle#35808)

  GuoxiaWang

  authored
  fcfb0af

  Copy full SHA for fcfb0af

• test=document_fix (PaddlePaddle#35824)

  GuoxiaWang

  authored
  177bf52

  Copy full SHA for 177bf52

• add launch doc (PaddlePaddle#35634)

  Show description for 5548061
  GuoxiaWang

  authored
  5548061

  Copy full SHA for 5548061

Commits on Sep 16, 2021

• support l2_normalize float16 (PaddlePaddle#35776)

  Show description for b666fd3
  GuoxiaWang

  authored
  b666fd3

  Copy full SHA for b666fd3

Commits on Sep 15, 2021

• fix dim check of class center sample (PaddlePaddle#35733)

  GuoxiaWang

  authored
  a957734

  Copy full SHA for a957734

Commits on Sep 13, 2021

• support hybrid parallel inference helper class (PaddlePaddle#35576)

  Show description for dc3c845
  GuoxiaWang

  authored
  dc3c845

  Copy full SHA for dc3c845

Pagination

Previous

Next