-
Notifications
You must be signed in to change notification settings - Fork 85
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Add integration tests for Distributing Cloud Tuner.
Fix a few issues: - Duplicate epoch reports to Oracle - Excessive cash discovery Info/Error log PiperOrigin-RevId: 339757281
- Loading branch information
1 parent
651ce41
commit a79009e
Showing
4 changed files
with
208 additions
and
20 deletions.
There are no files selected for viewing
161 changes: 161 additions & 0 deletions
161
...thon/tensorflow_cloud/tuner/tests/integration/distributing_cloudtuner_integration_test.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,161 @@ | ||
| # Lint as: python3 | ||
| # Copyright 2020 Google LLC. All Rights Reserved. | ||
| # | ||
| # Licensed under the Apache License, Version 2.0 (the "License"); | ||
| # you may not use this file except in compliance with the License. | ||
| # You may obtain a copy of the License at | ||
| # | ||
| # http://www.apache.org/licenses/LICENSE-2.0 | ||
| # | ||
| # Unless required by applicable law or agreed to in writing, software | ||
| # distributed under the License is distributed on an "AS IS" BASIS, | ||
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| # See the License for the specific language governing permissions and | ||
| # limitations under the License. | ||
| """Integration tests for Distributing Cloud Tuner.""" | ||
|
|
||
| import contextlib | ||
| import io | ||
| import os | ||
| import re | ||
| import kerastuner | ||
| import tensorflow as tf | ||
| from tensorflow import keras | ||
| from tensorflow_cloud.tuner import optimizer_client | ||
| from tensorflow_cloud.tuner.tuner import DistributingCloudTuner | ||
|
|
||
| # If input dataset is created outside tuner.search(), | ||
| # it requires eager execution even in TF 1.x. | ||
| if tf.version.VERSION.split(".")[0] == "1": | ||
| tf.compat.v1.enable_eager_execution() | ||
|
|
||
| # The project id to use to run tests. | ||
| _PROJECT_ID = os.environ["PROJECT_ID"] | ||
|
|
||
| # The GCP region in which the end-to-end test is run. | ||
| _REGION = os.environ["REGION"] | ||
|
|
||
| # Study ID for testing | ||
| _STUDY_ID_BASE = "dct_{}".format((os.environ["BUILD_ID"]).replace("-", "_")) | ||
|
|
||
| # The base docker image to use for the remote environment. | ||
| _DOCKER_IMAGE = os.environ["DOCKER_IMAGE"] | ||
|
|
||
| # The staging bucket to use to copy the model and data for the remote run. | ||
| _REMOTE_DIR = os.path.join("gs://", os.environ["TEST_BUCKET"], _STUDY_ID_BASE) | ||
|
|
||
| # The search space for hyperparameters | ||
| _HPS = kerastuner.engine.hyperparameters.HyperParameters() | ||
| _HPS.Float("learning_rate", min_value=1e-4, max_value=1e-2, sampling="log") | ||
| _HPS.Int("num_layers", 2, 10) | ||
|
|
||
|
|
||
| def _load_data(dir_path=None): | ||
| """Loads and prepares data.""" | ||
|
|
||
| mnist_file_path = None | ||
| if dir_path: | ||
| mnist_file_path = os.path.join(dir_path, "mnist.npz") | ||
|
|
||
| (x, y), (val_x, val_y) = keras.datasets.mnist.load_data(mnist_file_path) | ||
| x = x.astype("float32") / 255.0 | ||
| val_x = val_x.astype("float32") / 255.0 | ||
|
|
||
| return ((x[:10000], y[:10000]), (val_x, val_y)) | ||
|
|
||
|
|
||
| def _build_model(hparams): | ||
| # Note that CloudTuner does not support adding hyperparameters in | ||
| # the model building function. Instead, the search space is configured | ||
| # by passing a hyperparameters argument when instantiating (constructing) | ||
| # the tuner. | ||
| model = keras.Sequential() | ||
| model.add(keras.layers.Flatten(input_shape=(28, 28))) | ||
|
|
||
| # Build the model with number of layers from the hyperparameters | ||
| for _ in range(hparams.get("num_layers")): | ||
| model.add(keras.layers.Dense(units=64, activation="relu")) | ||
| model.add(keras.layers.Dense(10, activation="softmax")) | ||
|
|
||
| # Compile the model with learning rate from the hyperparameters | ||
| model.compile( | ||
| optimizer=keras.optimizers.Adam(lr=hparams.get("learning_rate")), | ||
| loss="sparse_categorical_crossentropy", | ||
| metrics=["acc"], | ||
| ) | ||
| return model | ||
|
|
||
|
|
||
| class _DistributingCloudTunerIntegrationTestBase(tf.test.TestCase): | ||
|
|
||
| def setUp(self): | ||
| super(_DistributingCloudTunerIntegrationTestBase, self).setUp() | ||
| self._study_id = None | ||
|
|
||
| def _assert_output(self, fn, regex_str): | ||
| stdout = io.StringIO() | ||
| with contextlib.redirect_stdout(stdout): | ||
| fn() | ||
| output = stdout.getvalue() | ||
| self.assertRegex(output, re.compile(regex_str, re.DOTALL)) | ||
|
|
||
| def _assert_results_summary(self, fn): | ||
| self._assert_output( | ||
| fn, ".*Results summary.*Trial summary.*Hyperparameters.*") | ||
|
|
||
| def _delete_dir(self, path) -> None: | ||
| """Deletes a directory if exists.""" | ||
| if tf.io.gfile.isdir(path): | ||
| tf.io.gfile.rmtree(path) | ||
|
|
||
| def tearDown(self): | ||
| super(_DistributingCloudTunerIntegrationTestBase, self).tearDown() | ||
|
|
||
| # Delete the study used in the test, if present | ||
| if self._study_id: | ||
| service = optimizer_client.create_or_load_study( | ||
| _PROJECT_ID, _REGION, self._study_id, None) | ||
| service.delete_study() | ||
|
|
||
| tf.keras.backend.clear_session() | ||
|
|
||
| # Delete log files, saved_models and other training assets | ||
| self._delete_dir(_REMOTE_DIR) | ||
|
|
||
|
|
||
| class DistributingCloudTunerIntegrationTest( | ||
| _DistributingCloudTunerIntegrationTestBase): | ||
|
|
||
| def setUp(self): | ||
| super(DistributingCloudTunerIntegrationTest, self).setUp() | ||
| (self._x, self._y), (self._val_x, self._val_y) = _load_data( | ||
| self.get_temp_dir()) | ||
|
|
||
| def testCloudTunerHyperparameters(self): | ||
| """Test case to configure Distributing Tuner with HyperParameters.""" | ||
| study_id = "{}_hyperparameters".format(_STUDY_ID_BASE) | ||
| self._study_id = study_id | ||
|
|
||
| tuner = DistributingCloudTuner( | ||
| _build_model, | ||
| project_id=_PROJECT_ID, | ||
| region=_REGION, | ||
| objective="acc", | ||
| hyperparameters=_HPS, | ||
| max_trials=2, | ||
| study_id=study_id, | ||
| directory=_REMOTE_DIR, | ||
| container_uri=_DOCKER_IMAGE | ||
| ) | ||
|
|
||
| tuner.search( | ||
| x=self._x, | ||
| y=self._y, | ||
| epochs=2, | ||
| validation_data=(self._val_x, self._val_y), | ||
| ) | ||
|
|
||
| self._assert_results_summary(tuner.results_summary) | ||
|
|
||
| if __name__ == "__main__": | ||
| tf.test.main() |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters