Commits

Breadcrumbs

History for

onmaster

User selector

All users

Datepicker

All time

Commit History

Commits on Jul 19, 2019

• Create datasets outside strategy.scope in tutorials and examples.

  Show description for 529a0e8
  

  guptapriya

  authored and

  copybara-github

  committed
  529a0e8

  Copy full SHA for 529a0e8

• Original cl/257106978 by priyag@. Just cleaning it up and removing num_gpu parameter.

  Show description for c35e1f6
  

  yashk2810

  authored and

  copybara-github

  committed
  c35e1f6

  Copy full SHA for c35e1f6

Commits on Jun 21, 2019

• Internal

  Show description for b991813
  

  yashk2810

  authored and

  copybara-github

  committed
  b991813

  Copy full SHA for b991813

Commits on Jun 18, 2019

• Add names so that checkpoint might recognize it

  Show description for 42d6507
  

  yashk2810

  authored and

  copybara-github

  committed
  42d6507

  Copy full SHA for 42d6507

• Switch division to tf.float32

  Show description for 8ca44dd
  

  yashk2810

  authored and

  copybara-github

  committed
  8ca44dd

  Copy full SHA for 8ca44dd

• Change to using manual loss calculation instead of using tf.nn.compute_average_loss

  Show description for 3e6c617
  

  yashk2810

  authored and

  copybara-github

  committed
  3e6c617

  Copy full SHA for 3e6c617

Commits on Jun 17, 2019

• Make pix2pix flexible so that cyclegan can use it.

  Show description for f28b124
  

  yashk2810

  authored and

  copybara-github

  committed
  f28b124

  Copy full SHA for f28b124

Commits on Jun 11, 2019

• Update densenet with the new DS api's.

  Show description for fc00b64
  

  yashk2810

  authored and

  copybara-github

  committed
  fc00b64

  Copy full SHA for fc00b64

Commits on May 5, 2019

• Fix loss scaling for densenet example. We were mistakenly scaling the entire loss by global batch size. Regularizer losses should not be scaled by batch size, only by number of replicas.

  Show description for 55e79ef
  

  guptapriya

  authored and

  copybara-github

  committed
  55e79ef

  Copy full SHA for 55e79ef

Commits on May 3, 2019

• Add validation of accuracy.

  tfboyd

  committed
  b8db1b0

  Copy full SHA for b8db1b0

Commits on Apr 19, 2019

• Update examples to account for changes to loss scaling in cl/243909950 and cl/244080246

  Show description for ff12b55
  

  guptapriya

  authored and

  copybara-github

  committed
  ff12b55

  Copy full SHA for ff12b55

Commits on Apr 18, 2019

• Switch to tf.image.resize_with_crop_or_pad

  Show description for 25d2da4
  

  yashk2810

  authored and

  copybara-github

  committed
  25d2da4

  Copy full SHA for 25d2da4

Commits on Apr 17, 2019

• Python 3 converts division to float. Converting local_batch_sz to int

  Show description for f8bf16f
  

  yashk2810

  authored and

  copybara-github

  committed
  f8bf16f

  Copy full SHA for f8bf16f

• Fix dec_input shape

  Show description for 223265c
  

  yashk2810

  authored and

  copybara-github

  committed
  223265c

  Copy full SHA for 223265c

• Fix distributed nmt bug

  Show description for 83b8a54
  

  yashk2810

  authored and

  copybara-github

  committed
  83b8a54

  Copy full SHA for 83b8a54

• Change to batch_size = 64

  Show description for ab49b0b
  

  yashk2810

  authored and

  copybara-github

  committed
  ab49b0b

  Copy full SHA for ab49b0b

Commits on Apr 3, 2019

• unicode_literals

  Show description for ba09c62
  

  MarkDaoust

  authored and

  copybara-github

  committed
  ba09c62

  Copy full SHA for ba09c62

Commits on Mar 26, 2019

• Adding nmt_with_attention

  Show description for cafe496
  

  yashk2810

  authored and

  copybara-github

  committed
  cafe496

  Copy full SHA for cafe496

Commits on Mar 4, 2019

• Adding a README

  Show description for 67f285c
  

  yashk2810

  authored and

  copybara-github

  committed
  67f285c

  Copy full SHA for 67f285c

Commits on Mar 1, 2019

• Densenet Cleanup

  Show description for abd5369
  

  yashk2810

  authored and

  copybara-github

  committed
  abd5369

  Copy full SHA for abd5369

• Making sure tf.image.resize is run in tests.

  Show description for 267b143
  

  yashk2810

  authored and

  copybara-github

  committed
  267b143

  Copy full SHA for 267b143

• Resize Image API was changed on 2/28 and this is a fix for that so that the benchmarks don't fail.

  Show description for 47b8498
  

  yashk2810

  authored and

  copybara-github

  committed
  47b8498

  Copy full SHA for 47b8498

Commits on Feb 27, 2019

• Proportionally increase batch_size so that we get maximum performance out of the GPU's and scale the learning rate too.

  Show description for da48328
  

  yashk2810

  authored and

  copybara-github

  committed
  da48328

  Copy full SHA for da48328

Commits on Feb 26, 2019

• Passing iterator as default instead of lambda

  Show description for e9c7a56
  

  yashk2810

  authored and

  copybara-github

  committed
  e9c7a56

  Copy full SHA for e9c7a56

• Densenet with Dist strat

  Show description for 35cb674
  

  yashk2810

  authored and

  copybara-github

  committed
  35cb674

  Copy full SHA for 35cb674

Commits on Feb 25, 2019

• Renaming to models

  Show description for cf642c3
  

  yashk2810

  authored and

  copybara-github

  committed
  cf642c3

  Copy full SHA for cf642c3

• .result() returns a tensor where .numpy() will return a value.

  Show description for 8be2848
  

  yashk2810

  authored and

  lamberta

  committed
  8be2848

  Copy full SHA for 8be2848

Commits on Feb 24, 2019

• Removing dropout because data augmentation is set

  Show description for 30db916
  

  yashk2810

  authored and

  lamberta

  committed
  30db916

  Copy full SHA for 30db916

Commits on Feb 23, 2019

• prefetch to improve performance

  Show description for 0a89b65
  

  yashk2810

  authored and

  lamberta

  committed
  0a89b65

  Copy full SHA for 0a89b65

• channels_first

  Show description for c2ab486
  

  yashk2810

  authored and

  lamberta

  committed
  c2ab486

  Copy full SHA for c2ab486

Commits on Feb 22, 2019

• Shifting to resize_image_with_crop_or_pad. Also shifting to channels_last to test if V100 is faster with that.

  Show description for 10fd5a4
  

  yashk2810

  authored and

  lamberta

  committed
  10fd5a4

  Copy full SHA for 10fd5a4

• Adding more data augmentation and using SGD optimizer with nesterov momentum.

  Show description for 6419595
  

  yashk2810

  authored and

  lamberta

  committed
  6419595

  Copy full SHA for 6419595

• Adding kwargs so that the benchmark framework won't return errors when extras is empty.

  Show description for b906e49
  

  yashk2810

  authored and

  lamberta

  committed
  b906e49

  Copy full SHA for b906e49

• switching to channels_first

  Show description for c2dc997
  

  yashk2810

  authored and

  lamberta

  committed
  c2dc997

  Copy full SHA for c2dc997

• Adding data augmentation

  Show description for 4d1e457
  

  yashk2810

  authored and

  lamberta

  committed
  4d1e457

  Copy full SHA for 4d1e457

Pagination

Previous

Next