| Commit message (Collapse) | Author | Age |
|
|
|
|
|
| |
shared resources are very similar to global variables functionally and they are initialized at the same time but since workers are only waiting for global variables being initialized, there is a race condition that sometimes the shared resource is not ready.
PiperOrigin-RevId: 216208679
|
|
|
|
|
|
| |
Fixes #21405.
PiperOrigin-RevId: 215973401
|
|
|
|
|
|
| |
Adds a bit of sanity checking by default to load_weights (e.g. for the case when absolutely nothing matches) while still supporting restore-on-create and the addition of new Layers to checkpointed models.
PiperOrigin-RevId: 215652168
|
|\
| |
| |
| | |
PiperOrigin-RevId: 215639962
|
| |
| |
| |
| |
| |
| | |
allows us to identify if we need to set the drop_remainder option when creating Dataset objects.
PiperOrigin-RevId: 215633097
|
| |
| |
| |
| |
| |
| | |
consume multiple mini-batches while some may not even one.
PiperOrigin-RevId: 215617588
|
| |
| |
| |
| | |
PiperOrigin-RevId: 215307701
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
output depends on the updates across all mirrors. Before this change,
update() would return a Mirrored value that where each component was
an update to a single mirror. This caused a problem since for reading
purposes other DistributionStrategy methods would consider it okay
to read any single component, and so if you for example did something
like session.run(strategy.update(...)) it would only perform the
update on one replica. The fix is to have the output be a Mirrored
value that is actually the identity operation returning the output on
that device, but that has a control dependency making sure that the
update actually happens on all the replicas. This fix was already
present in MirroredVariable._assign_func, this CL moves the fix into
update() and generalizes it to multiple return values.
To disable this new grouping behavior, you may now pass
"grouped=False" to update(). For example, some callers (like Optimizer)
are performing a lot of updates and they prefer to group all of them
together at once for performance reasons. In this case, we still want
to make sure the caller executes the update on all replicas, so we
return an unwrapped value instead of a Mirrored value. This has the
happy side effect of removing a bunch of unwrap calls in client code,
since unwrapping was the only safe way to use the Mirrored value we
used to return.
PiperOrigin-RevId: 215301909
|
| |
| |
| |
| | |
PiperOrigin-RevId: 215287936
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
https://github.com/tensorflow/community/pull/16.
In addition to the changes in the doc, I made the following updates (these changes make sense to me and I didn't notice them when compiling the doc):
* deprecate saved_model.builder.SavedModelBuilder - replaced with saved_model.SavedModelBuilder
* deprecate python_io.tf_record_iterator - replaced with io.tf_record_iterator
* deprecate python_io.TFRecordWriter - replaced with io.TFRecordWriter
* move reduce_join to tf.string
PiperOrigin-RevId: 215253944
|
| |
| |
| |
| | |
PiperOrigin-RevId: 215027511
|
|/ |
|
|
|
|
| |
PiperOrigin-RevId: 214824023
|
|
|
|
|
|
| |
mentions incorrect default value.
PiperOrigin-RevId: 214731772
|
|
|
|
| |
PiperOrigin-RevId: 214709465
|
|
|
|
|
|
| |
instead of internal processor objects.
PiperOrigin-RevId: 214678470
|
|
|
|
|
|
|
|
| |
objects in a tf.train.Saver() if var_list was a dict.
Includes the logic used for list in the dict code path.
PiperOrigin-RevId: 214324913
|
|
|
|
|
|
|
|
| |
self.test_session() has been deprecated in 9962eb5e84b15e309410071b06c2ed2d6148ed44 as its name confuses readers of the test. Moving to cached_session() instead which is more explicit about:
* the fact that the session may be reused.
* the session is not closed even when doing a "with self.test_session()" statement.
PiperOrigin-RevId: 214300210
|
|
|
|
|
|
| |
self.session(graph=...) as it's the same semantic.
PiperOrigin-RevId: 214286845
|
|
|
|
| |
PiperOrigin-RevId: 214263489
|
|
|
|
|
|
| |
gRPC failure and suggest increasing the number of parameter servers.
PiperOrigin-RevId: 214077622
|
|\
| |
| |
| | |
PiperOrigin-RevId: 213648091
|
| |
| |
| |
| |
| |
| | |
file, so that folks looking at API documentation can find the readme as well.
PiperOrigin-RevId: 213499832
|
| |
| |
| |
| |
| |
| |
| |
| | |
This enables cleanup of the variables referenced in defunned methods of objects when the object is garbage collected. Since one PolymorphicFunction is created per @defun, decorated methods before this change held on to all of the variables referenced in that method for any instance of the class (i.e. variables which should have been object-scoped were scoped to the lifetime of the class definition).
Raises an exception if variables used in the function have been deleted when it is called, which means no local variables.
PiperOrigin-RevId: 213337256
|
| | |
|
| |
| |
| |
| |
| |
| | |
of Optimizer.compute_gradients().
PiperOrigin-RevId: 213060585
|
| |
| |
| |
| |
| |
| |
| |
| | |
self.test_session() has been deprecated in 9962eb5e84b15e309410071b06c2ed2d6148ed44 as its name confuses readers of the test. Moving to cached_session() instead which is more explicit about:
* the fact that the session may be reused.
* the session is not closed even when doing a "with self.test_session()" statement.
PiperOrigin-RevId: 212336464
|
| |
| |
| |
| |
| |
| |
| |
| | |
self.test_session() has been deprecated in 9962eb5e84b15e309410071b06c2ed2d6148ed44 as its name confuses readers of the test. Moving to cached_session() instead which is more explicit about:
* the fact that the session may be reused.
* the session is not closed even when doing a "with self.test_session()" statement.
PiperOrigin-RevId: 212336352
|
| |
| |
| |
| |
| |
| | |
Also tests copying Checkpointable objects, which seems to just work.
PiperOrigin-RevId: 212289140
|
| |
| |
| |
| |
| |
| |
| |
| | |
Moves around some SaveableObjects to support a freeze method for python state saveables, and makes sure that the object graph proto is included in the frozen Saver.
This should be useful for embedding in SavedModels, where variables can be updated and the resulting checkpoints (saved from the SaverDef in the SavedModel) will still support Keras-style object-based restoration into Python programs (with better eager support and less fragile variable matching). This is also a step toward Estimators saving object-based checkpoints.
PiperOrigin-RevId: 212017296
|
| |
| |
| |
| |
| |
| | |
This is because many ops take longer during the first step due to autotune. Instead, the first timeline is now outputed after N seconds/steps.
PiperOrigin-RevId: 211854304
|
| |
| |
| |
| |
| |
| | |
These APIs are based on queue runners, which have been deprecated and will be removed in TensorFlow 2.0. They have been replaced with `tf.data.Dataset`, which provides a more efficient version of the same functionality.
PiperOrigin-RevId: 211727844
|
| |
| |
| |
| |
| |
| |
| |
| | |
TF 2.0 to return a no-arg callable to output a learning rate, instead of directly outputting a learning rate tensor.
This brings the graph mode API in line with the eager execution API, where this change was made to allow changing the learning rate value across different invocations of optimizer functions.
PiperOrigin-RevId: 211726295
|
| |
| |
| |
| |
| |
| | |
second axis of Tensors through tf.train.warm_start. Note that the underlying initializer already has this functionality (for example, for output layers).
PiperOrigin-RevId: 211709879
|
| |
| |
| |
| |
| |
| | |
These APIs are based on queue runners, which have been deprecated and will be removed in TensorFlow 2.0. They have been replaced with `tf.data.Dataset`, which provides a more efficient version of the same functionality.
PiperOrigin-RevId: 211700442
|
| |
| |
| |
| |
| |
| | |
If you want all-reduce, supply the `value` to the `destinations` argument.
PiperOrigin-RevId: 211148002
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
back rollback cl/210656847).
Fixing reference to _get_export_outputs_for_spec in TFMA (This function was refactored out, so the string has been removed from the list of methods that are copied from core Estimator).
*** Original change description ***
Automated rollback of commit 069f808e5c0462819bcd6c73c75491b00cdd42c2
PiperOrigin-RevId: 211122893
|
| |
| |
| |
| |
| |
| |
| |
| | |
runners defined.
This complements the deprecation warning that is printed when that function is called, and provides an actionable hint that the user can delete the call.
PiperOrigin-RevId: 211012334
|
| |
| |
| |
| |
| |
| | |
Queue runners will be removed in TensorFlow 2.0. They have been replaced with `tf.data` input pipelines, which provide a more efficient version of the same functionality.
PiperOrigin-RevId: 210964268
|
| |
| |
| |
| |
| |
| |
| | |
step counter. This allows us to get rid of the increment_var()
function and just use a standard assign_add().
PiperOrigin-RevId: 210743165
|
| |
| |
| |
| | |
PiperOrigin-RevId: 210656847
|
| |
| |
| |
| |
| | |
RELNOTES: tfdbg: Limit the total disk space occupied by dumped tensor data to 100 GBytes. Add environment variable `TFDBG_DISK_BYTES_LIMIT` to allow adjustment of this upper limit.
PiperOrigin-RevId: 210648585
|
| |
| |
| |
| |
| |
| | |
should not end up in the accumulator.
PiperOrigin-RevId: 210648271
|
| |
| |
| |
| | |
PiperOrigin-RevId: 210648154
|
| |
| |
| |
| |
| |
| | |
Restore previously checked if a key existed, but didn't quite ignore that value properly if it was missing.
PiperOrigin-RevId: 210455409
|
|\ \
| |/
|/|
| | |
PiperOrigin-RevId: 210392464
|
| |
| |
| |
| |
| |
| | |
Modifies the clock by an epsilon to fix creation and instant reloads of CheckpointState.
PiperOrigin-RevId: 210175050
|
| |
| |
| |
| |
| |
| | |
Doesn't delete anything. Also keeps paths to all checkpoints; I will follow up with a way to remove them manually.
PiperOrigin-RevId: 210128785
|
| |
| |
| |
| |
| |
| | |
The evaluate should be within the session scope.
PiperOrigin-RevId: 209983400
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
In the process removes unnecessary restriction on it being called multiple
times on the same variables; it might be necessary to do so if you want to
call ema.apply with different control dependencies in different calls to
session.run, for example.
PiperOrigin-RevId: 209945355
|