..
cluster_resolver
Make task_type and task_id standard properties in tf.distribute cluster resolvers.
2020-06-22 14:48:14 -07:00
experimental
model_collection
Remove passing experimental_run_tf_function in most tests.
2020-02-27 13:34:09 -08:00
parallel_device
Wrap save/restore logic in tf.function when in eager mode. This allows parallel saving and restoring when using multiple devices.
2020-06-22 13:23:14 -07:00
__init__.py
Expose MultiWorkerMirroredStrategy and ParameterServerStrategy
2019-02-20 17:57:26 -08:00
all_reduce_test.py
Fixed cases where tf.TensorShape was constructed with float dimensions
2019-10-14 20:53:35 -07:00
all_reduce.py
Merge pull request #36713 from tensorflow:terrytangyuan-patch-2
2020-02-13 10:26:24 -08:00
BUILD
Make cluster_resolver standard property in tf.distribute strategies.
2020-06-22 18:19:14 -07:00
central_storage_strategy.py
Drop experimental and v2 qualifiers from Strategy experimental_run_v2 method.
2020-03-12 10:26:26 -07:00
checkpoint_utils_test.py
Small adjustments on import spacing.
2019-12-18 20:32:12 -08:00
checkpointing_test.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
collective_all_reduce_strategy_test.py
Make cluster_resolver standard property in tf.distribute strategies.
2020-06-22 18:19:14 -07:00
collective_all_reduce_strategy.py
Make cluster_resolver standard property in tf.distribute strategies.
2020-06-22 18:19:14 -07:00
collective_util.py
Rename all_reduce_sum_gradients to experimental_aggregate_gradients
2020-03-24 13:31:15 -07:00
combinations_test.py
Add multi worker mirrored strategy combinations to enable easier testing again
2020-04-06 15:59:19 -07:00
combinations.py
Treat "test_xla_gpu" as GPU_TEST in "NamedGPUCombination".
2020-04-30 12:31:23 -07:00
cross_device_ops_test.py
Fix cross_device_ops_test with multi GPU
2020-06-16 15:33:47 -07:00
cross_device_ops.py
Explicitly take the set of devices in CollectiveAllReduce
2020-06-16 13:20:30 -07:00
cross_device_utils_test.py
Add an experimental_hints to batch all reduce
2020-02-25 20:32:44 -08:00
cross_device_utils.py
Explicitly take the set of devices in CollectiveAllReduce
2020-06-16 13:20:30 -07:00
ctl_correctness_test.py
Removing v1 optimizer tests from ctl_correctness_test to speed it up
2020-04-30 14:21:05 -07:00
custom_training_loop_gradient_test.py
Support Google-internal TPU resolution in strategy combinations.
2020-05-27 14:29:14 -07:00
custom_training_loop_input_test.py
Support packed variable for tf data captured function.
2020-06-22 16:24:37 -07:00
custom_training_loop_metrics_test.py
Drop experimental and v2 qualifiers from Strategy experimental_run_v2 method.
2020-03-12 10:26:26 -07:00
custom_training_loop_models_test.py
When calling strategy.reduce
in eager mode, wrap the strategy.run
calls inside with tf.function
so it is compatible with TPUStrategy.
2020-05-22 12:23:49 -07:00
custom_training_loop_optimizer_test.py
Error when experimental_aggregate_gradients=False is used with
2020-03-24 20:22:12 -07:00
device_util_test.py
Try to deduce job, replica and task from config.list_logical_devices() again
2020-06-16 15:22:24 -07:00
device_util.py
Try to deduce job, replica and task from config.list_logical_devices() again
2020-06-16 15:22:24 -07:00
distribute_config.py
distribute_coordinator_context.py
distribute_coordinator_test.py
minor spelling tweaks
2020-02-11 15:09:21 +09:00
distribute_coordinator.py
Distribute Coordinator currently assumes TF_CONFIG to be the only way to configure a strategy. We now allow cluster resolvers to be passed as arguments to instantiate the strategy instead of TF_CONFIG which should be used instead if set by the user.
2020-03-16 12:03:17 -07:00
distribute_lib_test.py
Make cluster_resolver standard property in tf.distribute strategies.
2020-06-22 18:19:14 -07:00
distribute_lib.py
Make cluster_resolver standard property in tf.distribute strategies.
2020-06-22 18:19:14 -07:00
distribute_utils.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
distributed_file_utils_test.py
Ensure distributed_file_utils.remove_temp_dirpath() can be safely called multiple times.
2020-04-27 16:05:05 -07:00
distributed_file_utils.py
Ensure distributed_file_utils.remove_temp_dirpath() can be safely called multiple times.
2020-04-27 16:05:05 -07:00
distribution_strategy_context.py
Drop experimental and v2 qualifiers from Strategy experimental_run_v2 method.
2020-03-12 10:26:26 -07:00
estimator_training.py
minor spelling tweaks
2020-02-11 15:09:21 +09:00
input_lib_test.py
Enable last partial batch for MWMS in TF2.x
2020-06-22 17:27:34 -07:00
input_lib_type_spec_test.py
Fix incompatibilities between DistributedIterator and the corresponding DistributedIteratorSpec.
2020-04-30 18:10:48 -07:00
input_lib.py
Enable last partial batch for MWMS in TF2.x
2020-06-22 17:27:34 -07:00
input_ops_test.py
input_ops.py
keras_metrics_test.py
Drop experimental and v2 qualifiers from Strategy experimental_run_v2 method.
2020-03-12 10:26:26 -07:00
keras_save_load_test.py
Remove passing experimental_run_tf_function in most tests.
2020-02-27 13:34:09 -08:00
metrics_v1_test.py
Move to using 'initializer' from 'initialize' to be more consistent with the tf.data APIs.
2020-01-15 13:20:27 -08:00
minimize_loss_test.py
Drop experimental and v2 qualifiers from Strategy experimental_run_v2 method.
2020-03-12 10:26:26 -07:00
mirrored_function_strategy_test.py
Drop experimental and v2 qualifiers from Strategy experimental_run_v2 method.
2020-03-12 10:26:26 -07:00
mirrored_function_strategy.py
Another round of refactoring of values.py to split utility functions that use distributed Variable types defined in values.py.
2020-06-12 12:04:39 -07:00
mirrored_run.py
Another round of refactoring of values.py to split utility functions that use distributed Variable types defined in values.py.
2020-06-12 12:04:39 -07:00
mirrored_strategy_test.py
Fork the keras related mirrored_strategy_test to keras/distribute.
2020-06-19 10:27:02 -07:00
mirrored_strategy.py
Graduate TPUStrategy from experimental.
2020-06-20 13:10:50 -07:00
mirrored_variable_test.py
Fork keras related mirror_variable_test to keras/distribute.
2020-06-19 10:57:00 -07:00
model_combinations.py
moving_averages_test.py
In assign_moving_average
, call update_fn
instead of strategy.extended.update(var, update_fn)
when in update context.
2020-06-10 17:52:55 -07:00
multi_process_lib.py
Set TF_FORCE_GPU_ALLOW_GROWTH for multi process tests
2020-06-12 12:22:38 -07:00
multi_process_runner_no_init_test.py
Refactor MultiProcessRunner:
2019-11-08 20:24:05 -08:00
multi_process_runner_test.py
Fix tsan failure in multi_process_runner_test.
2020-06-22 15:50:17 -07:00
multi_process_runner.py
Fix tsan failure in multi_process_runner_test.
2020-06-22 15:50:17 -07:00
multi_worker_continuous_run_test.py
Improve multi_process_runner
2020-06-11 05:20:30 -07:00
multi_worker_test_base.py
Mark multi-process utilities with subprocess module as deprecated in favor of using MultiProcessRunner.
2020-05-11 15:07:56 -07:00
multi_worker_util_test.py
Allow evaluator not in cluster_spec, to be consistent with legacy Estimator.
2019-11-21 15:40:54 -08:00
multi_worker_util.py
Allow evaluator not in cluster_spec, to be consistent with legacy Estimator.
2019-11-21 15:40:54 -08:00
numpy_dataset_test.py
Add tf.distribute.Strategy.experimental_make_numpy_iterator()
function.
2019-01-09 14:10:49 -08:00
numpy_dataset.py
Remove *args from disribute/ variable creators
2020-02-01 16:58:14 -08:00
one_device_strategy_test.py
Create a DistributedDataset/DistributedDatasetV1 based on TF2/1.x mode.
2020-04-14 01:06:49 -07:00
one_device_strategy.py
Add InputOptions to experimental_distribute_dataset(s_from_function).
2020-06-15 16:05:58 -07:00
packed_distributed_variable_test.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
packed_distributed_variable.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
parameter_server_strategy_test.py
Another round of refactoring of values.py to split utility functions that use distributed Variable types defined in values.py.
2020-06-12 12:04:39 -07:00
parameter_server_strategy.py
Add InputOptions to experimental_distribute_dataset(s_from_function).
2020-06-15 16:05:58 -07:00
ps_values_test.py
Refactor values.py into a utility file and a PS values file.
2020-05-28 10:58:13 -07:00
ps_values.py
Override "map_resources" in AggregatingVariable.
2020-06-19 16:18:14 -07:00
README.md
Graduate TPUStrategy from experimental.
2020-06-20 13:10:50 -07:00
reduce_util.py
minor spelling tweaks
2020-02-11 15:09:21 +09:00
remote_mirrored_strategy_eager_test.py
Support LogicalDevice in MirroredStrategy config
2019-11-13 15:19:24 -08:00
saved_model_mixed_api_test.py
Migrate saved_model_mixed_api_test to V2 API
2020-06-11 10:50:30 -07:00
saved_model_save_load_test.py
Drop experimental and v2 qualifiers from Strategy experimental_run_v2 method.
2020-03-12 10:26:26 -07:00
saved_model_test_base.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
sharded_variable_test.py
Add ShardedVariable
class.
2019-10-03 15:05:15 -07:00
sharded_variable.py
Support ShardedVariable in tf.keras.layers.Embedding
.
2020-05-21 11:06:33 -07:00
shared_variable_creator_test.py
shared_variable_creator.py
Remove *args from disribute/ variable creators
2020-02-01 16:58:14 -08:00
single_loss_example.py
step_fn_test.py
step_fn.py
Move to using 'initializer' from 'initialize' to be more consistent with the tf.data APIs.
2020-01-15 13:20:27 -08:00
strategy_combinations_test.py
Drop experimental and v2 qualifiers from Strategy experimental_run_v2 method.
2020-03-12 10:26:26 -07:00
strategy_combinations.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
strategy_common_test.py
Make cluster_resolver standard property in tf.distribute strategies.
2020-06-22 18:19:14 -07:00
strategy_reduce_test.py
When calling strategy.reduce
in eager mode, wrap the strategy.run
calls inside with tf.function
so it is compatible with TPUStrategy.
2020-05-22 12:23:49 -07:00
strategy_test_lib.py
Another round of refactoring of values.py to split utility functions that use distributed Variable types defined in values.py.
2020-06-12 12:04:39 -07:00
summary_op_util.py
tf_function_test.py
Use first worker as default device in tf_function_test.
2020-06-08 11:32:39 -07:00
tpu_strategy_test.py
Support dynamic outputs for XLA on demand ops.
2020-07-01 11:11:47 -07:00
tpu_strategy.py
Make cluster_resolver standard property in tf.distribute strategies.
2020-06-22 18:19:14 -07:00
tpu_values.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
values_test.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
values_util.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
values.py
Support packed variable in DistributedVariable. Add an option to enable packed variable in TPUStrategy.
2020-06-18 20:12:02 -07:00
warm_starting_util_test.py
Small adjustments on import spacing.
2019-12-18 20:32:12 -08:00
zero_batch_test.py
Fix input size used for batch normalization.
2020-04-09 22:01:21 -07:00