In [ ]:

    

# Only for when running on Colab:
import sys
if 'google.colab' in sys.modules:
    # Get the dependency .py files, if any.
    ! git clone https://github.com/GoogleCloudPlatform/cloudml-samples.git
    ! cp cloudml-samples/tpu/templates/tpu_lstm_keras/* .

    # Authenticate the user for better GCS access.
    # Copy verification code into the text field to continue.
    from google.colab import auth
    auth.authenticate_user()


    

In [ ]:

    

import argparse
import os
import numpy as np
import tensorflow as tf
from tensorflow.keras.layers import LSTM


    

In [ ]:

    

def build_model():
    inputs = tf.keras.Input(shape=(5, 3))
    encoded = tf.keras.layers.LSTM(10)(inputs)
    outputs = tf.keras.layers.Dense(1, activation=tf.nn.sigmoid)(encoded)

    model = tf.keras.Model(inputs=inputs, outputs=outputs)

    return model


    

In [ ]:

    

def train_input_fn():
    batch_size = 16

    # make some fake data
    x = np.random.rand(100, 5, 3)
    y = np.random.rand(100, 1)

    # TPUs currently do not support float64
    x_tensor = tf.constant(x, dtype=tf.float32)
    y_tensor = tf.constant(y, dtype=tf.float32)

    # create tf.data.Dataset
    dataset = tf.data.Dataset.from_tensor_slices((x_tensor, y_tensor))

    dataset = dataset.repeat().shuffle(32).batch(batch_size, drop_remainder=True)

    # TPUs need to know all dimensions when the graph is built
    # Datasets know the batch size only when the graph is run
    def set_shapes(features, labels):
        features_shape = features.get_shape().merge_with([batch_size, None, None])
        labels_shape = labels.get_shape().merge_with([batch_size, None])

        features.set_shape(features_shape)
        labels.set_shape(labels_shape)

        return features, labels

    dataset = dataset.map(set_shapes)
    dataset = dataset.prefetch(tf.contrib.data.AUTOTUNE)

    return dataset


    

In [ ]:

    

def main(args):
    model = build_model()

    if args.use_tpu:
        # distribute over TPU cores
        # Note: This requires TensorFlow 1.11
        tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(args.tpu)
        strategy = tf.contrib.tpu.TPUDistributionStrategy(tpu_cluster_resolver)
        model = tf.contrib.tpu.keras_to_tpu_model(
            model, strategy=strategy)

    optimizer = tf.train.RMSPropOptimizer(learning_rate=0.05)
    loss_fn = tf.losses.log_loss
    model.compile(optimizer, loss_fn)

    model.fit(train_input_fn, epochs=3, steps_per_epoch=10)

    if not os.path.exists(args.model_dir):
        os.makedirs(args.model_dir)
    model.save(os.path.join(args.model_dir, 'model.hd5'))


    

In [ ]:

    

parser = argparse.ArgumentParser()

parser.add_argument(
    '--model-dir',
    type=str,
    default='/tmp/tpu-template',
    help='Location to write checkpoints and summaries to.  Must be a GCS URI when using Cloud TPU.')
parser.add_argument(
    '--use-tpu',
    action='store_true',
    help='Whether to use TPU.')
parser.add_argument(
    '--tpu',
    default=None,
    help='The name or GRPC URL of the TPU node.  Leave it as `None` when training on AI Platform.')

args, _ = parser.parse_known_args()


    

In [ ]:

    

# colab.research.google.com specific
if 'google.colab' in sys.modules:
    import json
    import os

    # TODO(user): change this
    args.model_dir = 'gs://your-gcs-bucket'

    # When connected to the TPU runtime
    if 'COLAB_TPU_ADDR' in os.environ:
        tpu_grpc = 'grpc://{}'.format(os.environ['COLAB_TPU_ADDR'])

        args.tpu = tpu_grpc
        args.use_tpu = True

        # Upload credentials to the TPU
        with tf.Session(tpu_grpc) as sess:
            data = json.load(open('/content/adc.json'))
            tf.contrib.cloud.configure_gcs(sess, credentials=data)


    

In [ ]:

    

main(args)