Commits

Breadcrumbs

History for

onmaster

User selector

All users

Datepicker

All time

Commit History

Commits on Oct 22, 2019

• Update XLNet classifier:

  Show description for 444e8c7
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  444e8c7

  Copy full SHA for 444e8c7

• Adds save_steps to fix squad

  Show description for 9616954
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  9616954

  Copy full SHA for 9616954

Commits on Oct 21, 2019

• Pass the file patterns instead of flattened file paths into tf.data.Dataset.list_file. This helps Cloud DF4x4 Bert Startup from 11min34s to 6min53s.

  Show description for ad5a14f
  

  rxsang

  authored and

  tensorflower-gardener

  committed
  ad5a14f

  Copy full SHA for ad5a14f

• Consolidate DistributeOptions.auto_shard into DistributeOptions.auto_shard_policy.

  Show description for bd0d569
  

  Frank Chen

  authored and

  tensorflower-gardener

  committed
  bd0d569

  Copy full SHA for bd0d569

Commits on Oct 19, 2019

• Always pass **kwargs to __call__ override for custom layers.

  Show description for d169c20
  

  rachellj218

  authored and

  tensorflower-gardener

  committed
  d169c20

  Copy full SHA for d169c20

• Fix regression

  Show description for 9ac5487
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  9ac5487

  Copy full SHA for 9ac5487

• Internal change

  Show description for 49ba237
  tensorflower-gardener

  committed
  49ba237

  Copy full SHA for 49ba237

Commits on Oct 18, 2019

• Fix xlnet save_steps and steps_per_epoch conflicts

  Show description for e6750c5
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  e6750c5

  Copy full SHA for e6750c5

• Support online masking for XLNet

  Show description for 6cd426d
  

  rachellj218

  authored and

  tensorflower-gardener

  committed
  6cd426d

  Copy full SHA for 6cd426d

Commits on Oct 17, 2019

• Makes do_lower_case not trainable.

  Show description for eabff6f
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  eabff6f

  Copy full SHA for eabff6f

Commits on Oct 16, 2019

• Do not scale loss manually for BERT classifier when compile/fit() API is used.

  Show description for 306e3e1
  tensorflower-gardener

  committed
  306e3e1

  Copy full SHA for 306e3e1

Commits on Oct 15, 2019

• Export the pretrained transformer-xl model for finetuning stage.

  Show description for 048f5a9
  

  rachellj218

  authored and

  tensorflower-gardener

  committed
  048f5a9

  Copy full SHA for 048f5a9

• Adds assets for bert export

  Show description for 188536e
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  188536e

  Copy full SHA for 188536e

• Fix variable mismatch between XLNet pretrain and finetune tasks.

  Show description for ddd45b8
  

  rachellj218

  authored and

  tensorflower-gardener

  committed
  ddd45b8

  Copy full SHA for ddd45b8

Commits on Oct 13, 2019

• Internal change

  Show description for dc4e7c4
  tensorflower-gardener

  committed
  dc4e7c4

  Copy full SHA for dc4e7c4

• Internal change

  Show description for 9a833e2
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  9a833e2

  Copy full SHA for 9a833e2

Commits on Oct 11, 2019

• Change summary directory and model checkpoint directory so that training via Keras Compile/Fit() and custom training loop is consistent.

  Show description for e67a206
  tensorflower-gardener

  committed
  e67a206

  Copy full SHA for e67a206

• Internal changes.

  Show description for ad1a37c
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  ad1a37c

  Copy full SHA for ad1a37c

• Bert Hub usage for squad

  Show description for 223f514
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  223f514

  Copy full SHA for 223f514

Commits on Oct 9, 2019

• Internal change

  Show description for 7207422
  tensorflower-gardener

  committed
  7207422

  Copy full SHA for 7207422

Commits on Oct 7, 2019

• Bert fp16 perf improvements, do the matmul in intermediate later in fp16, and also remove explicit casting to fp32 for layerNorm.

  Show description for 3fa6560
  tensorflower-gardener

  committed
  3fa6560

  Copy full SHA for 3fa6560

Commits on Oct 4, 2019

• Internal change

  Show description for b045ce7
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  b045ce7

  Copy full SHA for b045ce7

Commits on Oct 3, 2019

• Merge pull request tensorflow#7616 from Huawei-MRC-OSI:fix-bert-messages

  Show description for 0f176f6
  tensorflower-gardener

  committed
  0f176f6

  Copy full SHA for 0f176f6

• Adds Bert tf2 tfhub usage and export script.

  Show description for 73fcb8f
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  73fcb8f

  Copy full SHA for 73fcb8f

• Fix bert_modeling error messages

  

  Sergey Mironov

  committed
  da2d13b

  Copy full SHA for da2d13b

Commits on Sep 27, 2019

• Support variable reshape to make TF1 checkpoint compatible with a Bert without reshape in einsum layers.

  Show description for f56c495
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  f56c495

  Copy full SHA for f56c495

Commits on Sep 24, 2019

• Use experimental_connect_to_cluster API in TPU lib to support training on a slice of a TPU pod.

  Show description for 497989e
  

  Bruce Fontaine

  authored and

  tensorflower-gardener

  committed
  497989e

  Copy full SHA for 497989e

• Perfzero XLNet classifier Imdb accuracy test on 8 GPUs.

  Show description for a52564c
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  a52564c

  Copy full SHA for a52564c

Commits on Sep 20, 2019

• Update readme

  Show description for d03ce00
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  d03ce00

  Copy full SHA for d03ce00

• Internal change

  Show description for a1fca62
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  a1fca62

  Copy full SHA for a1fca62

Commits on Sep 18, 2019

• Update readme

  Show description for c4bacbf
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  c4bacbf

  Copy full SHA for c4bacbf

• doc, make clearer statement about fine-tuning scripts (tensorflow#7572)

  

  Sergey Mironov

  committed
  0f02d68

  Copy full SHA for 0f02d68

Commits on Sep 17, 2019

• Fix squad script.

  Show description for f66efa5
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  f66efa5

  Copy full SHA for f66efa5

• Move Bert to NLP. Tasks are moved to nlp/bert/

  Show description for 1862b9c
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  1862b9c

  Copy full SHA for 1862b9c

Commits on Sep 13, 2019

• Fix xlnet imports

  Show description for c21bec5
  

  saberkun

  authored and

  tensorflower-gardener

  committed
  c21bec5

  Copy full SHA for c21bec5

Pagination

Previous

Next