Added the "Keras - Train classifier" component (#809)

* Added sample component

* Replaced human-readable names with pythonic names

Some people are confused.

* Removed the tag from image name.

components/sample/keras/train_classifier/Dockerfile

@@ -0,0 +1,5 @@
+ARG BASE_IMAGE_TAG=1.12.0-py3
+FROM tensorflow/tensorflow:$BASE_IMAGE_TAG
+RUN python3 -m pip install keras
+COPY ./src /pipelines/component/src
+ENTRYPOINT python3 /pipelines/component/src/train.py

components/sample/keras/train_classifier/README.md

@@ -0,0 +1,20 @@
+## Train classification model using Keras ##
+
+Usage:
+
+~~~~
+#Load the component
+train_op = comp.load_component(url='https://raw.githubusercontent.com/Ark-kun/pipelines/Added-sample-component/components/sample/keras/train_classifier/component.yaml')
+
+#Use the component as part of the pipeline
+def pipeline():
+    train_task = train_op(
+        training_set_features_path=os.path.join(testdata_root, 'training_set_features.tsv'),
+        training_set_labels_path=os.path.join(testdata_root, 'training_set_labels.tsv'),
+        output_model_uri=os.path.join(temp_dir_name, 'outputs/output_model/data'),
+        model_config=Path(testdata_root).joinpath('model_config.json').read_text(),
+        number_of_classes=2,
+        number_of_epochs=10,
+        batch_size=32,
+    )
+~~~~

components/sample/keras/train_classifier/build_image.sh

@@ -0,0 +1,31 @@
+#!/bin/bash -e
+# Copyright 2018 Google LLC
+#
+# 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.
+
+image_name=gcr.io/ml-pipeline/sample/keras/train_classifier
+image_tag=latest
+full_image_name=${image_name}:${image_tag}
+base_image_tag=1.12.0-py3
+
+cd "$(dirname "$0")"
+
+docker build --build-arg BASE_IMAGE_TAG=$base_image_tag -t "$full_image_name" .
+docker push "$full_image_name"
+
+#Output the strict image name (which contains the sha256 image digest)
+#This name can be used by the subsequent steps to refer to the exact image that was built even if another image with the same name was pushed.
+image_name_with_digest=$(docker inspect --format="{{index .RepoDigests 0}}" "$IMAGE_NAME")
+strict_image_name_output_file=./versions/image_digests_for_tags/$image_tag
+mkdir -p "$(dirname "$strict_image_name_output_file")"
+echo $image_name_with_digest | tee "$strict_image_name_output_file"

components/sample/keras/train_classifier/component.yaml

@@ -0,0 +1,27 @@
+name: Keras - Train classifier
+description: Trains classifier using Keras sequential model
+inputs:
+  - {name: training_set_features_path, type: {GcsUri: TSV}, description: 'Local or GCS path to the training set features table.'}
+  - {name: training_set_labels_path, type: {GcsUri: TSV}, description: 'Local or GCS path to the training set labels (each label is a class index from 0 to num-classes - 1).'}
+  - {name: output_model_uri, type: {GcsUri: Keras model}, description: 'Local or GCS path specifying where to save the trained model. The model (topology + weights + optimizer state) is saved in HDF5 format and can be loaded back by calling keras.models.load_model'} #Remove GcsUri and move to outputs once artifact passing support is checked in.
+  - {name: model_config, type: {GcsUri: Keras model config json}, description: 'JSON string containing the serialized model structure. Can be obtained by calling model.to_json() on a Keras model.'}
+  - {name: number_of_classes, type: Integer, description: 'Number of classifier classes.'}
+  - {name: number_of_epochs, type: Integer, default: '100', description: 'Number of epochs to train the model. An epoch is an iteration over the entire `x` and `y` data provided.'}
+  - {name: batch_size, type: Integer, default: '32', description: 'Number of samples per gradient update.'}
+outputs:
+  - {name: output_model_uri, type: {GcsUri: Keras model}, description: 'GCS path where the trained model has been saved. The model (topology + weights + optimizer state) is saved in HDF5 format and can be loaded back by calling keras.models.load_model'} #Remove GcsUri and make it a proper output once artifact passing support is checked in.
+implementation:
+  container:
+    image: gcr.io/ml-pipeline/sample/keras/train_classifier
+    command: [python3, /pipelines/component/src/train.py]
+    args: [
+      --training-set-features-path, {inputValue: training_set_features_path},
+      --training-set-labels-path, {inputValue: training_set_labels_path},
+      --output-model-path, {inputValue: output_model_uri},
+      --model-config-json, {inputValue: model_config},
+      --num-classes, {inputValue: number_of_classes},
+      --num-epochs, {inputValue: number_of_epochs},
+      --batch-size, {inputValue: batch_size},
+
+      --output-model-path-file, {outputPath: output_model_uri},
+    ]

components/sample/keras/train_classifier/run_tests.sh

@@ -0,0 +1,17 @@
+#!/bin/bash -e
+# Copyright 2018 Google LLC
+#
+# 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.
+
+cd $(dirname $0)
+python3 -m unittest discover --verbose --start-dir tests --top-level-directory=..

components/sample/keras/train_classifier/src/train.py

@@ -0,0 +1,79 @@
+# Copyright 2018 Google LLC
+#
+# 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.
+
+import argparse
+import json
+import os
+from pathlib import Path
+
+import keras
+import numpy as np
+
+parser = argparse.ArgumentParser(description='Train classifier model using Keras')
+
+parser.add_argument('--training-set-features-path', type=str, help='Local or GCS path to the training set features table.')
+parser.add_argument('--training-set-labels-path', type=str, help='Local or GCS path to the training set labels (each label is a class index from 0 to num-classes - 1).')
+parser.add_argument('--output-model-path', type=str, help='Local or GCS path specifying where to save the trained model. The model (topology + weights + optimizer state) is saved in HDF5 format and can be loaded back by calling keras.models.load_model')
+parser.add_argument('--model-config-json', type=str, help='JSON string containing the serialized model structure. Can be obtained by calling model.to_json() on a Keras model.')
+parser.add_argument('--num-classes', type=int, help='Number of classifier classes.')
+parser.add_argument('--num-epochs', type=int, default=100, help='Number of epochs to train the model. An epoch is an iteration over the entire `x` and `y` data provided.')
+parser.add_argument('--batch-size', type=int, default=32, help='Number of samples per gradient update.')
+
+parser.add_argument('--output-model-path-file', type=str, help='Path to a local file containing the output model URI. Needed for data passing until the artifact support is checked in.') #TODO: Remove after the team agrees to let me check in artifact support.
+args = parser.parse_args()
+
+# The data, split between train and test sets:
+#(x_train, y_train), (x_test, y_test) = cifar10.load_data()
+x_train = np.loadtxt(args.training_set_features_path)
+y_train = np.loadtxt(args.training_set_labels_path)
+print('x_train shape:', x_train.shape)
+print(x_train.shape[0], 'train samples')
+
+# Convert class vectors to binary class matrices.
+y_train = keras.utils.to_categorical(y_train, args.num_classes)
+
+model = keras.models.model_from_json(args.model_config_json)
+
+model.add(keras.layers.Activation('softmax'))
+
+# initiate RMSprop optimizer
+opt = keras.optimizers.rmsprop(lr=0.0001, decay=1e-6)
+
+# Let's train the model using RMSprop
+model.compile(loss='categorical_crossentropy',
+              optimizer=opt,
+              metrics=['accuracy'])
+
+x_train = x_train.astype('float32')
+x_train /= 255
+
+model.fit(
+    x_train,
+    y_train,
+    batch_size=args.batch_size,
+    epochs=args.num_epochs,
+    shuffle=True
+)
+
+# Save model and weights
+if not args.output_model_path.startswith('gs://'):
+    save_dir = os.path.dirname(args.output_model_path)
+    if not os.path.isdir(save_dir):
+        os.makedirs(save_dir)
+
+model.save(args.output_model_path)
+print('Saved trained model at %s ' % args.output_model_path)
+
+Path(args.output_model_path_file).parent.mkdir(parents=True, exist_ok=True)
+Path(args.output_model_path_file).write_text(args.output_model_path)

components/sample/keras/train_classifier/tests/test_component.py

@@ -0,0 +1,64 @@
+# Copyright 2018 Google LLC
+#
+# 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.
+
+import os
+import subprocess
+import tempfile
+import unittest
+from contextlib import contextmanager
+from pathlib import Path
+
+import kfp.components as comp
+
+@contextmanager
+def components_local_output_dir_context(output_dir: str):
+    old_dir = comp._components._outputs_dir
+    try:
+        comp._components._outputs_dir = output_dir
+        yield output_dir
+    finally:
+        comp._components._outputs_dir = old_dir
+
+class KerasTrainClassifierTestCase(unittest.TestCase):
+    def test_handle_training_xor(self):
+        tests_root = os.path.abspath(os.path.dirname(__file__))
+        component_root = os.path.abspath(os.path.join(tests_root, '..'))
+        testdata_root = os.path.abspath(os.path.join(tests_root, 'testdata'))
+
+        train_op = comp.load_component(os.path.join(component_root, 'component.yaml'))
+
+        with tempfile.TemporaryDirectory() as temp_dir_name:
+            with components_local_output_dir_context(temp_dir_name):
+                train_task = train_op(
+                    training_set_features_path=os.path.join(testdata_root, 'training_set_features.tsv'),
+                    training_set_labels_path=os.path.join(testdata_root, 'training_set_labels.tsv'),
+                    output_model_uri=os.path.join(temp_dir_name, 'outputs/output_model/data'),
+                    model_config=Path(testdata_root).joinpath('model_config.json').read_text(),
+                    number_of_classes=2,
+                    number_of_epochs=10,
+                    batch_size=32,
+                )
+
+            full_command = train_task.command + train_task.arguments
+            full_command[0] = 'python'
+            full_command[1] = os.path.join(component_root, 'src', 'train.py')
+
+            process = subprocess.run(full_command)
+
+            (output_model_uri_file, ) = (train_task.file_outputs['output-model-uri'], )
+            output_model_uri = Path(output_model_uri_file).read_text()
+
+
+if __name__ == '__main__':
+    unittest.main()

components/sample/keras/train_classifier/tests/testdata/model_config.json

@@ -0,0 +1,66 @@
+{
+    "class_name": "Sequential",
+    "config": {
+        "name": "sequential_1",
+        "layers": [
+            {
+                "class_name": "Dense",
+                "config": {
+                    "name": "dense_1",
+                    "trainable": true,
+                    "units": 2,
+                    "activation": "linear",
+                    "use_bias": true,
+                    "kernel_initializer": {
+                        "class_name": "VarianceScaling",
+                        "config": {
+                            "scale": 1.0,
+                            "mode": "fan_avg",
+                            "distribution": "uniform",
+                            "seed": null
+                        }
+                    },
+                    "bias_initializer": {
+                        "class_name": "Zeros",
+                        "config": {}
+                    },
+                    "kernel_regularizer": null,
+                    "bias_regularizer": null,
+                    "activity_regularizer": null,
+                    "kernel_constraint": null,
+                    "bias_constraint": null
+                }
+            },
+            {
+                "class_name": "Dense",
+                "config": {
+                    "name": "dense_2",
+                    "trainable": true,
+                    "units": 2,
+                    "activation": "linear",
+                    "use_bias": true,
+                    "kernel_initializer": {
+                        "class_name": "VarianceScaling",
+                        "config": {
+                            "scale": 1.0,
+                            "mode": "fan_avg",
+                            "distribution": "uniform",
+                            "seed": null
+                        }
+                    },
+                    "bias_initializer": {
+                        "class_name": "Zeros",
+                        "config": {}
+                    },
+                    "kernel_regularizer": null,
+                    "bias_regularizer": null,
+                    "activity_regularizer": null,
+                    "kernel_constraint": null,
+                    "bias_constraint": null
+                }
+            }
+        ]
+    },
+    "keras_version": "2.2.4",
+    "backend": "tensorflow"
+}

components/sample/keras/train_classifier/tests/testdata/training_set_features.tsv

@@ -0,0 +1,4 @@
+0 0
+0 1
+1 0
+1 1

components/sample/keras/train_classifier/tests/testdata/training_set_labels.tsv

@@ -0,0 +1,4 @@
+0
+1
+1
+0