def test_basic_logging(capsys):
with valohai.logger() as logger:
logger.log("myint", 123)
captured = capsys.readouterr()
assert captured.out == '\n{"myint": 123}\n'
with valohai.logger() as logger:
logger.log("I am float", 0.1241234435877)
captured = capsys.readouterr()
assert captured.out == '\n{"I am float": 0.1241234435877}\n'
def test_partial_logging(capsys):
with valohai.logger() as logger:
logger.log("myint", 123)
logger.log("myfloat", 0.5435)
captured = capsys.readouterr()
assert captured.out == '\n{"myint": 123, "myfloat": 0.5435}\n'
logger = valohai.logger()
logger.log("myint", 234)
logger.log("myfloat", 0.2322323)
logger.log("extrathing", 2)
logger.flush()
captured = capsys.readouterr()
assert captured.out == '\n{"myint": 234, "myfloat": 0.2322323, "extrathing": 2}\n'
def log(epoch, logs):
with vh.logger() as logger:
logger.log('epoch', epoch)
logger.log('accuracy', logs['accuracy'])
logger.log('loss', logs['loss'])
# This enables Valohai to version your metadata
# and for you to use it to compare experiments
def log(epoch, logs):
with vh.logger() as logger:
logger.log('epoch', epoch)
logger.log('accuracy', logs['accuracy'])
logger.log('loss', logs['loss'])
cb = tf.keras.callbacks.LambdaCallback(on_epoch_end=log)
model.fit(x_train,
y_train,
epochs=vh.parameters('epochs').value,
callbacks=[cb])
# Evaluate the model and print out the test metrics as JSON
test_loss, test_acc = model.evaluate(x_test, y_test, verbose=2)
with vh.logger() as logger:
logger.log('test_accuracy', test_acc)
logger.log('test_loss', test_loss)
# Write output files to Valohai outputs directory
# This enables Valohai to version your data
# and upload output it to the default data store
model.save(vh.outputs('model').path('model.h5'))
def log_metadata(epoch, logs):
"""Helper function to log training metrics"""
with valohai.logger() as logger:
logger.log('epoch', epoch)
logger.log('accuracy', logs['accuracy'])
logger.log('loss', logs['loss'])
def main():
# valohai.prepare enables us to update the valohai.yaml configuration file with
# the Valohai command-line client by running `valohai yaml step train_model.py`
valohai.prepare(
step='train-model',
image='tensorflow/tensorflow:2.6.0',
default_inputs={
'dataset':
'https://valohaidemo.blob.core.windows.net/mnist/preprocessed_mnist.npz',
},
default_parameters={
'learning_rate': 0.001,
'epochs': 5,
},
)
# Read input files from Valohai inputs directory
# This enables Valohai to version your training data
# and cache the data for quick experimentation
input_path = valohai.inputs('dataset').path()
with np.load(input_path, allow_pickle=True) as f:
x_train, y_train = f['x_train'], f['y_train']
x_test, y_test = f['x_test'], f['y_test']
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(28, 28)),
tf.keras.layers.Dense(128, activation='relu'),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(10),
])
optimizer = tf.keras.optimizers.Adam(
learning_rate=valohai.parameters('learning_rate').value)
loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)
model.compile(optimizer=optimizer, loss=loss_fn, metrics=['accuracy'])
# Print metrics out as JSON
# This enables Valohai to version your metadata
# and for you to use it to compare experiments
callback = tf.keras.callbacks.LambdaCallback(on_epoch_end=log_metadata)
model.fit(x_train,
y_train,
epochs=valohai.parameters('epochs').value,
callbacks=[callback])
# Evaluate the model and print out the test metrics as JSON
test_loss, test_accuracy = model.evaluate(x_test, y_test, verbose=2)
with valohai.logger() as logger:
logger.log('test_accuracy', test_accuracy)
logger.log('test_loss', test_loss)
# Write output files to Valohai outputs directory
# This enables Valohai to version your data
# and upload output it to the default data store
suffix = uuid.uuid4()
output_path = valohai.outputs().path(f'model-{suffix}.h5')
model.save(output_path)