Skip to content
This repository has been archived by the owner on Jul 7, 2023. It is now read-only.

v1.3 #449

Merged
merged 33 commits into from
Nov 29, 2017
Merged

v1.3 #449

merged 33 commits into from
Nov 29, 2017

Conversation

rsepassi
Copy link
Contributor

No description provided.

Ryan Sepassi and others added 30 commits November 29, 2017 13:34
More tiny sketch fixes
6cf47f9 
PiperOrigin-RevId: 175621997
Add Modality.targets_weights_fn
50f5515 
PiperOrigin-RevId: 175722118
Update layout optimizer config code. pruning and constfold are enable…
176efe6 
…d by default now (no need to explicitly specify them any more).

PiperOrigin-RevId: 175743440
Preserve static shape info where possible with shape_list and enable …
e0a2f86 
…constant ones_matrix_band_part

PiperOrigin-RevId: 175748936
Fix weights_fn calls
957a384 
PiperOrigin-RevId: 175864073
Refactor rev_block into a RevBlock class that exposes forward and bac…
8cacb79 
…kward

PiperOrigin-RevId: 175878950
Clean up revblock code with tf.contrib.framework.nest
976da09 
PiperOrigin-RevId: 175902336
Add ImageImagenet224; resnet50 now fits batch size of 128
717afe9 
PiperOrigin-RevId: 175904903
Added an option to omit quotes from the subtoken voc format.
2e6d512 
PiperOrigin-RevId: 175908007
Bug fix, make class lab a VarLen feature to be compatible with img2im…
dccc9ac 
…g datasets that don't have labels.

PiperOrigin-RevId: 175994994
Update tpu_trainer to use T2TModel.model_fn and upate RunConfig for n…
7fe103f 
…on-tpu use

PiperOrigin-RevId: 176058540
Fix SymbolModality weights fn
8ca96b8 
PiperOrigin-RevId: 176127650
Add support for multi-gpu training with tpu_trainer
c0ce3dd 
PiperOrigin-RevId: 176169403
CHECKPOINT BREAKING: make T2TModel a subclass of Layer so it can be c…
01b8c31 
…alled; all variables are now in model-name scope.

PiperOrigin-RevId: 176407831
Remove unused functions from transformer_vae.
2145729 
PiperOrigin-RevId: 176584099
Update distance computation for k-nearest neighbours to be more effic…
f77da80 
…ient, by computing the norms separately.

PiperOrigin-RevId: 176585527
@nshazeer
Pass the modified hparams to the modalities, so they can know the mode.
cc80721 
PiperOrigin-RevId: 176688435
@nshazeer
This change breaks previous checkpoints. Make Transformer fast on TPU.
c10e016 
PiperOrigin-RevId: 176747359
Changes to make t2t tf.eager compatible
b104292 
PiperOrigin-RevId: 176783794
@nshazeer
This change breaks previous checkpoints. Make Transformer fast on TPU.
b3cad0c 
PiperOrigin-RevId: 176784764
@nshazeer
Versions of problems with combined examples. Good for TPU training.
936db05 
PiperOrigin-RevId: 176807931
Small bug fix.
8e1958e 
PiperOrigin-RevId: 177058606
When using eager, use slow decoding for transformer models.
676e272 
PiperOrigin-RevId: 177062937
merge PRs.
6d9b5e1 
PiperOrigin-RevId: 177069364
Small updates to T2TModel.__call__ callers
a3d0ffe 
PiperOrigin-RevId: 177073383
Make daisy_chain_variables and hparam instead of flag and unset to al…
5adacd0 
…low LSTM to train in distributed mode.

PiperOrigin-RevId: 177193238
Move Estimator input_fn and model_fn construction into Problem and T2…
398e85b 
…TModel, respectively, which allows subclassing

PiperOrigin-RevId: 177229237
@nshazeer
This change breaks previous checkpoints. Make Transformer fast on TPU.
0ffe0e6 
PiperOrigin-RevId: 177255666
Correction for eager-mode decoding scopes (to use with pre-trained ch…
8f05bab 
…eckpoints).

PiperOrigin-RevId: 177261645
Make default Parallelism in T2TModel reusing (removes "parallel_0" fr…
2ca3232 
…om variable names in colab).

PiperOrigin-RevId: 177265031
nshazeer and others added 3 commits November 29, 2017 13:37
@nshazeer
fix bug with bpe32k - prepend <pad> and <EOS> to vocab.
7909c69 
PiperOrigin-RevId: 177271941
Disable image summary in eager mode (tf.Eager doesn't like it for now).
c65a646 
PiperOrigin-RevId: 177309768
v1.3
69701e4 
PiperOrigin-RevId: 177359986
@googlebot
Copy link

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for the commit author(s). If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and have the pull request author add another comment and the bot will run again.

lukaszkaiser
lukaszkaiser approved these changes Nov 29, 2017
View reviewed changes
Copy link
Contributor

@lukaszkaiser lukaszkaiser left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks!

@lukaszkaiser lukaszkaiser merged commit e3cd447 into tensorflow:master Nov 29, 2017
This was referenced Jan 19, 2018
Open
Open
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Reviewers

@lukaszkaiser lukaszkaiser lukaszkaiser approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@rsepassi @googlebot @lukaszkaiser @nshazeer