def train_and_evaluate(input_path: str,
model_dir: str,
num_epochs: int = 5,
batch_size: int = 128,
learning_rate: float = 0.01) -> None:
"""Trains and evaluates the Keras model.
Uses the Keras model defined in model.py. Saves the trained model in TensorFlow SavedModel
format to the path defined in part by the --job-dir argument."""
# Split datasets into training and testing
train_feature, eval_feature, train_target, eval_target = utils.load_data(
input_path)
# [END ai_platform_tfkeras_task_train_and_evaluate_load]
# [START ai_platform_tfkeras_task_train_and_evaluate_dimensions]
# Extract dimensions of the data
num_train_examples, input_dim = train_feature.shape
num_eval_examples = eval_feature.shape[1]
output_dim = train_target.shape[1]
# [END ai_platform_tfkeras_task_train_and_evaluate_dimensions]
# [START ai_platform_tfkeras_task_train_and_evaluate_model]
# Create the Keras Model
keras_model = model.create_keras_model(
input_dim=input_dim,
output_dim=output_dim,
learning_rate=learning_rate,
)
# [END ai_platform_tfkeras_task_train_and_evaluate_model]
# [START ai_platform_tfkeras_task_train_and_evaluate_training_data]
# Pass a numpy array by passing DataFrame.values
training_dataset = model.input_fn(
features=train_feature.values,
labels=train_target.values,
shuffle=True,
num_epochs=num_epochs,
batch_size=batch_size,
)
# [END ai_platform_tfkeras_task_train_and_evaluate_training_data]
# [START ai_platform_tfkeras_task_train_and_evaluate_validation_data]
# Pass a numpy array by passing DataFrame.values
validation_dataset = model.input_fn(
features=eval_feature.values,
labels=eval_target.values,
shuffle=False,
num_epochs=num_epochs,
batch_size=num_eval_examples,
)
# [END ai_platform_tfkeras_task_train_and_evaluate_validation_data]
# [START ai_platform_tfkeras_task_train_and_evaluate_fit_export]
# Train model
keras_model.fit(
training_dataset,
steps_per_epoch=int(num_train_examples / batch_size),
epochs=num_epochs,
validation_data=validation_dataset,
validation_steps=1,
verbose=1,
)
# Export model
keras_model.save(model_dir)
print(f"Model exported to: {model_dir}")
def train_and_evaluate(input_path: str,
job_dir: str,
num_epochs: int = 5,
batch_size: int = 128,
learning_rate: float = 0.01) -> None:
"""Trains and evaluates the Keras model.
Uses the Keras model defined in model.py. Saves the trained model in TensorFlow SavedModel
format to the path defined in part by the --job-dir argument."""
# Split datasets into training and testing
train_feature, eval_feature, train_target, eval_target = utils.load_data(
input_path)
# [END ai_platform_tfkeras_task_train_and_evaluate_load]
# [START ai_platform_tfkeras_task_train_and_evaluate_dimensions]
# Extract dimensions of the data
num_train_examples, input_dim = train_feature.shape
num_eval_examples = eval_feature.shape[1]
output_dim = train_target.shape[1]
# [END ai_platform_tfkeras_task_train_and_evaluate_dimensions]
# [START ai_platform_tfkeras_task_train_and_evaluate_model]
# Create the Keras Model
keras_model = model.create_keras_model(
input_dim=input_dim,
output_dim=output_dim,
learning_rate=learning_rate,
)
# [END ai_platform_tfkeras_task_train_and_evaluate_model]
# [START ai_platform_tfkeras_task_train_and_evaluate_training_data]
# Pass a numpy array by passing DataFrame.values
training_dataset = model.input_fn(
features=train_feature.values,
labels=train_target.values,
shuffle=True,
num_epochs=num_epochs,
batch_size=batch_size,
)
# [END ai_platform_tfkeras_task_train_and_evaluate_training_data]
# [START ai_platform_tfkeras_task_train_and_evaluate_validation_data]
# Pass a numpy array by passing DataFrame.values
validation_dataset = model.input_fn(
features=eval_feature.values,
labels=eval_target.values,
shuffle=False,
num_epochs=num_epochs,
batch_size=num_eval_examples,
)
# [END ai_platform_tfkeras_task_train_and_evaluate_validation_data]
# [START ai_platform_tfkeras_task_train_and_evaluate_tensorboard]
# Setup Learning Rate decay.
lr_decay_cb = tf.keras.callbacks.LearningRateScheduler(
lambda epoch: learning_rate + 0.02 * (0.5**(1 + epoch)), verbose=True)
# Setup TensorBoard callback.
tensorboard_cb = tf.keras.callbacks.TensorBoard(os.path.join(
job_dir, "keras_tensorboard"),
histogram_freq=1)
# [END ai_platform_tfkeras_task_train_and_evaluate_tensorboard]
# [START ai_platform_tfkeras_task_train_and_evaluate_fit_export]
# Train model
keras_model.fit(
training_dataset,
steps_per_epoch=int(num_train_examples / batch_size),
epochs=num_epochs,
validation_data=validation_dataset,
validation_steps=1,
verbose=1,
callbacks=[lr_decay_cb, tensorboard_cb],
)
# Export model
export_path = os.path.join(job_dir, "tfkeras_model/")
tf.keras.models.save_model(keras_model, export_path)
print(f"Model exported to: {export_path}")