| /external/tensorflow/tensorflow/python/distribute/ |
| D | mirrored_strategy_test.py | 70 distribution=[
80 def testMinimizeLoss(self, distribution): argument
82 self._test_minimize_loss_eager(distribution)
84 self._test_minimize_loss_graph(distribution)
86 def testReplicaId(self, distribution): argument
87 self._test_replica_id(distribution)
89 def testNumReplicasInSync(self, distribution): argument
90 self.assertEqual(2, distribution.num_replicas_in_sync)
92 def testCallAndMergeExceptions(self, distribution): argument
93 self._test_call_and_merge_exceptions(distribution)
[all …]
|
| D | custom_training_loop_input_test.py | 81 distribution=strategy_combinations.all_strategies,
84 def testConstantNumpyInput(self, distribution): argument
92 outputs = distribution.experimental_local_results(
93 distribution.run(computation, args=(x,)))
97 constant_op.constant(4., shape=(distribution.num_replicas_in_sync)),
102 distribution=strategy_combinations.all_strategies,
105 def testStatefulExperimentalRunAlwaysExecute(self, distribution): argument
106 with distribution.scope():
116 distribution.run(assign_add)
124 distribution=strategy_combinations.strategies_minus_tpu,
[all …]
|
| D | one_device_strategy_test.py | 34 distribution=[
43 def testMinimizeLoss(self, distribution): argument
45 self._test_minimize_loss_eager(distribution)
47 self._test_minimize_loss_graph(distribution)
49 def testReplicaId(self, distribution): argument
50 self._test_replica_id(distribution)
52 def testCallAndMergeExceptions(self, distribution): argument
53 self._test_call_and_merge_exceptions(distribution)
55 def testReplicateDataset(self, distribution): argument
65 self._test_input_fn_iterable(distribution, input_fn, expected_values)
[all …]
|
| D | mirrored_variable_test.py | 62 distribution=[
102 def testVariableInFuncGraph(self, distribution): argument
109 with func_graph.FuncGraph("fg").as_default(), distribution.scope():
111 v2 = distribution.extended.call_for_each_replica(model_fn)
113 self._test_mv_properties(v1, "foo:0", distribution)
114 self._test_mv_properties(v2, "bar:0", distribution)
116 def testVariableWithTensorInitialValueInFunction(self, distribution): argument
131 return distribution.experimental_local_results(
132 distribution.extended.call_for_each_replica(model_fn))
136 def testSingleVariable(self, distribution): argument
[all …]
|
| D | metrics_v1_test.py | 77 distribution=[
88 distribution=[
99 def _test_metric(self, distribution, dataset_fn, metric_fn, expected_fn): argument
100 with ops.Graph().as_default(), distribution.scope():
101 iterator = distribution.make_input_fn_iterator(lambda _: dataset_fn())
102 if isinstance(distribution, (tpu_strategy.TPUStrategy,
105 value, update = distribution.extended.call_for_each_replica(
108 return distribution.group(update)
110 ctx = distribution.extended.experimental_run_steps_on_iterator(
111 step_fn, iterator, iterations=distribution.extended.steps_per_run)
[all …]
|
| D | moving_averages_test.py | 48 distribution=all_distributions, mode=["graph"])
51 distribution=all_distributions, mode=["eager"], use_function=[True, False])
57 def testReplicaModeWithoutZeroDebias(self, distribution): argument
69 with distribution.scope():
70 var, assign = distribution.extended.call_for_each_replica(replica_fn)
73 self.evaluate(distribution.experimental_local_results(assign))
84 def testReplicaMode(self, distribution): argument
95 with distribution.scope():
96 var, assign_op = distribution.extended.call_for_each_replica(replica_fn)
99 self.evaluate(distribution.experimental_local_results(assign_op))
[all …]
|
| D | vars_test.py | 57 distribution=[
63 distribution=[
74 distribution=[
86 def testAssign(self, distribution, experimental_run_tf_function): argument
95 return distribution.experimental_local_results(
96 distribution.run(update_fn))
114 with distribution.scope():
126 def testAssignOnWriteVar(self, distribution, experimental_run_tf_function): argument
128 with distribution.scope():
141 return distribution.experimental_local_results(
[all …]
|
| D | input_lib_test.py | 290 distribution=[
296 def testDisablingOwnedIteratorsInTF2(self, distribution, input_type): argument
310 distribution)
314 distribution)
323 distribution._enable_legacy_iterators = True
332 distribution=[
335 def testMultiDeviceIterInitialize(self, distribution): argument
345 dataset_fn(distribute_lib.InputContext()), input_workers, distribution)
361 distribution=[
366 def testOneDeviceCPU(self, input_type, api_type, iteration_type, distribution, argument
[all …]
|
| D | values_test.py | 85 def _make_mirrored(distribution=None): argument
87 if distribution:
88 devices = distribution.extended.worker_devices
97 if (distribution is not None) and isinstance(distribution, _TPU_STRATEGIES):
101 mirrored = var_cls(distribution, v, variable_scope.VariableAggregation.SUM)
107 distribution=[
119 distribution=(strategy_combinations.all_strategies_minus_default +
123 def testMakeDistributedValueFromTensor(self, distribution): argument
132 distribution.experimental_distribute_values_from_function(value_fn))
134 ds_test_util.gather(distribution, distributed_values),
[all …]
|
| D | input_lib_type_spec_test.py | 55 distribution=[
60 def testTypeSpec(self, input_type, distribution, argument
67 distribution.extended.experimental_enable_get_next_as_optional = (
70 dist_dataset = distribution.experimental_distribute_dataset(dataset)
71 with distribution.scope():
87 distribution=[
93 distribution, enable_get_next_as_optional): argument
100 distribution.extended.experimental_enable_get_next_as_optional = (
103 dist_dataset = distribution.experimental_distribute_dataset(dataset)
104 with distribution.scope():
[all …]
|
| D | tf_function_test.py | 48 distribution=strategy_combinations.all_strategies,
53 self, distribution, run_functions_eagerly): argument
57 worker = distribution.extended.worker_devices[0]
62 with distribution.scope():
74 distribution=strategy_combinations.all_strategies,
79 self, distribution, run_functions_eagerly): argument
83 worker = distribution.extended.worker_devices[0]
88 with distribution.scope():
105 distribution=strategy_combinations.all_strategies,
109 def testReadVariableInsideFunction(self, distribution, run_functions_eagerly): argument
[all …]
|
| D | remote_mirrored_strategy_eager_test.py | 40 distribution=[
55 def testNumReplicasInSync(self, distribution): argument
56 self._testNumReplicasInSync(distribution)
58 def testMinimizeLoss(self, distribution): argument
59 self._testMinimizeLoss(distribution)
61 def testDeviceScope(self, distribution): argument
62 self._testDeviceScope(distribution)
64 def testMakeInputFnIteratorWithDataset(self, distribution): argument
65 self._testMakeInputFnIteratorWithDataset(distribution)
67 def testMakeInputFnIteratorWithCallable(self, distribution): argument
[all …]
|
| /external/bcc/tools/ |
| D | runqlat_example.txt | 12 usecs : count distribution
30 The distribution is bimodal, with one mode between 0 and 15 microseconds,
32 spikes in the ASCII distribution (which is merely a visual representation
49 msecs : count distribution
57 msecs : count distribution
65 msecs : count distribution
73 This shows a similar distribution across the three summaries.
82 msecs : count distribution
90 msecs : count distribution
98 msecs : count distribution
[all …]
|
| /external/tensorflow/tensorflow/python/ops/ |
| D | nn_loss_scaling_utilities_test.py | 49 distribution=[
53 def testComputeAverageLossDefaultGlobalBatchSize(self, distribution): argument
60 with distribution.scope():
61 per_replica_losses = distribution.run(
63 loss = distribution.reduce("SUM", per_replica_losses, axis=None)
68 distribution=[
72 def testComputeAverageLossSampleWeights(self, distribution): argument
73 with distribution.scope():
75 per_replica_losses = distribution.run(
79 loss = distribution.reduce("SUM", per_replica_losses, axis=None)
[all …]
|
| /external/tensorflow/tensorflow/python/keras/distribute/ |
| D | keras_dnn_correctness_test.py | 38 distribution=strategy_combinations.all_strategies,
40 distribution=strategy_combinations.multi_worker_mirrored_strategies,
46 distribution=keras_correctness_test_base.all_strategies,
62 distribution=None, argument
64 with keras_correctness_test_base.MaybeDistributionScope(distribution):
110 def test_dnn_correctness(self, distribution, use_numpy, use_validation_data): argument
111 self.run_correctness_test(distribution, use_numpy, use_validation_data)
117 def test_dnn_correctness_with_partial_last_batch_eval(self, distribution, argument
121 distribution, use_numpy, use_validation_data, partial_last_batch='eval')
127 def test_dnn_correctness_with_partial_last_batch(self, distribution, argument
[all …]
|
| D | keras_utils_test.py | 82 def test_callbacks_in_fit(self, distribution): argument
83 with distribution.scope():
90 dataset = keras_test_lib.get_dataset(distribution)
106 if (isinstance(distribution, tpu_strategy.TPUStrategyV1) and
110 steps_per_run = distribution.extended.steps_per_run
136 def test_callbacks_in_eval(self, distribution): argument
137 with distribution.scope():
144 dataset = keras_test_lib.get_dataset(distribution)
160 def test_callbacks_in_predict(self, distribution): argument
161 with distribution.scope():
[all …]
|
| D | minimize_loss_test.py | 84 distribution=[strategy_combinations.tpu_strategy],
88 distribution=[strategy_combinations.tpu_strategy],
92 def testTrainNetwork(self, distribution, optimizer_fn, use_callable_loss): argument
93 with distribution.scope():
100 return distribution.group(
101 distribution.extended.call_for_each_replica(
104 iterator = self._get_iterator(distribution, dataset_fn)
107 return distribution.extended.experimental_run_steps_on_iterator(
134 def testTrainNetworkByCallForEachReplica(self, distribution, optimizer_fn, argument
136 with distribution.scope():
[all …]
|
| D | custom_training_loop_models_test.py | 61 distribution=(strategy_combinations.all_strategies +
68 def test_single_keras_layer_run(self, distribution): argument
70 input_iterator = iter(distribution.experimental_distribute_dataset(dataset))
72 with distribution.scope():
85 outputs = distribution.run(
87 return nest.map_structure(distribution.experimental_local_results,
92 def test_keras_model_optimizer_run(self, distribution): argument
94 input_iterator = iter(distribution.experimental_distribute_dataset(dataset))
96 with distribution.scope():
111 outputs = distribution.run(step_fn, args=(replicated_inputs,))
[all …]
|
| D | saved_model_save_load_test.py | 49 distribution, argument
53 return test_base.load_and_run_with_saved_model_api(distribution, saved_dir,
59 distribution): argument
61 model_and_input, distribution)
67 distribution, save_in_scope): argument
69 model_and_input, distribution, save_in_scope)
86 def test_no_variable_device_placement(self, model_and_input, distribution, argument
88 saved_dir = self.run_test_save_strategy(model_and_input, distribution,
107 def _predict_with_model(self, distribution, model, predict_dataset): argument
108 if distribution:
[all …]
|
| D | keras_correctness_test_base.py | 62 combinations.combine(distribution=all_strategies),
68 combinations.combine(distribution=all_strategies),
74 combinations.combine(distribution=strategies_minus_tpu),
99 distribution=strategies_for_embedding_models()),
102 distribution=eager_mode_strategies),
112 combinations.combine(distribution=tpu_strategies),
118 combinations.combine(distribution=multi_worker_mirrored_strategies),
124 combinations.combine(distribution=multi_worker_mirrored_strategies),
131 def __init__(self, distribution): argument
132 self._distribution = distribution
[all …]
|
| D | distribute_strategy_test.py | 167 def batch_wrapper(dataset, batch_size, distribution, repeat=None): argument
172 if backend.is_tpu_strategy(distribution):
194 def get_dataset(distribution): argument
199 dataset = batch_wrapper(dataset, 10, distribution)
203 def get_predict_dataset(distribution): argument
207 dataset = batch_wrapper(dataset, 10, distribution)
241 distribution=strategies_minus_tpu, mode=['graph', 'eager'])
246 distribution=tpu_strategies, mode=['graph', 'eager'])
250 return combinations.combine(distribution=tpu_strategies, mode=['graph'])
255 distribution=multi_worker_mirrored_strategies, mode=['eager'])
[all …]
|
| D | custom_training_loop_metrics_test.py | 39 distribution=strategy_combinations.all_strategies +
43 def test_multiple_keras_metrics_experimental_run(self, distribution): argument
44 with distribution.scope():
55 distribution.run(step_fn)
64 distribution=strategy_combinations.all_strategies+
68 def test_update_keras_metric_declared_in_strategy_scope(self, distribution): argument
69 with distribution.scope():
73 dataset = distribution.experimental_distribute_dataset(dataset)
80 distribution.run(step_fn, args=(i,))
88 distribution=strategy_combinations.all_strategies,
[all …]
|
| D | saved_model_test_base.py | 77 distribution=strategies,
92 distribution=strategies,
104 def load_and_run_with_saved_model_api(distribution, saved_dir, predict_dataset, argument
108 if distribution:
109 dist_predict_dataset = distribution.experimental_distribute_dataset(
112 result = distribution.run(
118 reduced = distribution.experimental_local_results(result)
147 distribution, argument
176 def _predict_with_model(self, distribution, model, predict_dataset): argument
186 distribution): argument
[all …]
|
| D | custom_training_loop_optimizer_test.py | 39 distribution=keras_strategy_combinations.multidevice_strategies,
49 def test_custom_aggregation(self, distribution, argument
52 with distribution.scope():
69 return distribution.experimental_local_results(
70 distribution.run(step_fn, args=(grads,)))
76 distribution=strategy_combinations.one_device_strategy,
79 def test_custom_aggregation_one_device(self, distribution, argument
82 with distribution.scope():
96 return distribution.experimental_local_results(
97 distribution.run(step_fn, args=(grads,)))
[all …]
|
| /external/tensorflow/tensorflow/tools/api/golden/v1/ |
| D | tensorflow.distributions.pbtxt | 5 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
9 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
13 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
17 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
21 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
25 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
29 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
33 mtype: "<class \'tensorflow.python.ops.distributions.distribution.ReparameterizationType\'>"
37 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
41 mtype: "<class \'tensorflow.python.ops.distributions.distribution._DistributionMeta\'>"
[all …]
|