GitLab

* Merged commit includes the following changes:
249776315  by tianlin<tianlin@google.com>:

    Internal change

249763206  by tianlin<tianlin@google.com>:

    For TF 2.0 (related to Beam Search), expand cond dims in tf.where(cond, x, y) to make all parameters broadcastable.

--
249392724  by hongkuny<hongkuny@google.com>:

    Internal change

PiperOrigin-RevId: 249776315

* Merged commit includes the following changes:
249823043  by tianlin<tianlin@google.com>:

    Bring back v2 test for predict and eval.

--

PiperOrigin-RevId: 249823043

* Add a test enabling get_next_as_optional behavior.

* Remove repeated flag.

* Remove trailing space.

* Make the name shorter.

* Fix lint error.

* Refine the benchmark name.

Adding validation every epoch allows us to view the progress during training instead of having to wait until the last eval. Mostly useful for manual runs.

Current batch size 160000 does not converge to the desired HR. So we decrease to 99k which is known to converge. Tested locally and got to 63.5 at epoch 7. Also decreasing number of epochs as I don't see any improvement after epoch 7-8.

249580533  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Internal change

249566870  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Set up BERT benchmark test.

--

PiperOrigin-RevId: 249580533

* Add enable_get_next_as_optional flag.

* Set enable_get_next_as_optional to strategy.

* Add comments to explain the flag.

* Remove trailing whitespace.

* Remove trailing space.

249500988  by hongkuny<hongkuny@google.com>:

    Lints

--

PiperOrigin-RevId: 249500988

* Add big tests.

* fix super

* Add fp16, increase 8xGPU batch-sizes

* Adding the rest of the fp16 tests.

* Big accuracy test batch_perf_gpu

* fix docstrings

* add _run_and_report

* Edited docstrings

* Merged commit includes the following changes:
249218656  by tianlin<tianlin@google.com>:

    Deal with imports, fix a typo and make unit tests fast.

--
249198645  by tianlin<tianlin@google.com>:

    Trivial: Remove one empty line before "import tensorflow"

--
249195490  by tianlin<tianlin@google.com>:

    Initialize Transformer TF V2 Model with Keras subclassing implementation. (Compatible with TF V1)

--
249195008  by tianlin<tianlin@google.com>:

    Internal change

249173564  by hongkuny<hongkuny@google.com>:

    Internal change

249079258  by hongkuny<hongkuny@google.com>:

    Internal change

247691534  by haoyuzhang<haoyuzhang@google.com>:

    Internal change

247533725  by haoyuzhang<haoyuzhang@google.com>:

    Internal change

247509295  by haoyuzhang<haoyuzhang@google.com>:

    Internal change

247311355  by wangtz<wangtz@google.com>:

    Internal change

247303127  by wangtz<wangtz@google.com>:

  ...

249377254  by hongkuny<hongkuny@google.com>:

    Internal change

249373328  by hongkuny<hongkuny@google.com>:

    Clean up tf import

--
249333938  by hongkuny<hongkuny@google.com>:

    Fix tf1 import

--
249325089  by hongkuny<hongkuny@google.com>:

    BERT 2.0

--
249173564  by hongkuny<hongkuny@google.com>:

    Internal change

PiperOrigin-RevId: 249377254

* Delete accuracy if exists in eval results.

* get global_step only if it exists in eval results

This will allow one to easily reproduce a benchmark by running with the flags.

* Set the --clone_model_in_keras_dist_strat to None.  Remove the separate no_cloning benchmarks and add a couple of cloning ones.  Fixes the learning rate schedule to cache its ops per graph.

* Added 'tfdata_exp' version of all benchmarks which set
FLAGS.tf_data_experimental_slack = True. Renamed
`data_prefetch_with_slack` to `data_delay_prefetch` (haoyu's change)
to make the names more distinct.

* Add flag to resnet input pipeline and surface through
keras_imagenet_main.py

- Test passes locally python3 and test is already
    skipped for python2.

* Add FP16 and benchmarks.

* add missing run and report.

* Add loss_scale as option not included with dtype.

* move loss_scale validation under dtype conditional.

* add loss_scale to flags tested.

* Fix trivial model to work properly with fp16

* Add comment on manual casting

Previously we had one dense layer in trivial model. The weight was [224*224*3, num_classes].
Using two dense layers, the weights are [224*224*3, 1] and [1, num_classes].

* Do not report metrics in performance benchmarks

* Rename flag

* Modified tweaked tests to use tensor learning rate

* Add first benchmark and return stats.

* Remove print statements update training steps.

* Revert print T: in print statement.

* Remove print(stats)

* add 2 gpu accuracy test for base.

* Fixed total_batch_size when using gpu + gFile deprecations.

* 8 GPU test name fix

* Add 4 and 8 GPU tests.

* typo fixes.

* Clean up test names and methods.

* bleu uncased.  docstring format fix.

* Add learning rate tensor. This makes training slower

* Improve LearningRateSchedule with better efficiency

* Fix lint error

* Replace constant definition with existing one

* Enable CuDNN BatchNorm spatial persistent by default; Remove 2nd zero padding layer

* Apply scale=False and fused=True consistently to BatchNorm layers

* Undo remove padding layer

* Replace zero padding with padding attribute in max pooling for better performance

* Resolve comments

* Revert "Replace zero padding with padding attribute in max pooling for better performance"

This reverts commit ad49db057c800ecac008eec1057005bd2c08ac73.

This options allows the new tf.train.experimental.enable_mixed_precision_graph_rewrite() function to be used for fp16, instead of manual casts.

Replace per_device with per_replica and PerDevice with PerReplica, because the PerDevice concept was renamed and doesn't exist anymore. (#6693)

* Replace per_device with per_replica and PerDevice with PerReplica, because the PerReplica concept was renamed and doesn't exist anymore.