def train(training_set,
validation_set,
test_set,
model_definition,
save_path='model',
model_load_path=None,
resume=False,
skip_save_model=False,
skip_save_progress=False,
skip_save_log=False,
gpus=None,
gpu_fraction=1.0,
use_horovod=False,
random_seed=default_random_seed,
debug=False):
"""
:param training_set: Dataset contaning training data
:type training_set: Dataset
:param validation_set: Dataset contaning validation data
:type validation_set: Datasetk
:param test_set: Dataset contaning test data.
:type test_set: Dataset
:param model_definition: Model definition which defines the different
parameters of the model, features, preprocessing and training.
:type model_definition: Dictionary
:param save_path: The path to save the model to.
:type save_path: filepath (str)
:param model_load_path: If this is specified the loaded model will be used
as initialization (useful for transfer learning).
:type model_load_path: filepath (str)
:param skip_save_model: Disables
saving model weights and hyperparameters each time the model
improves. By default Ludwig saves model weights after each epoch
the validation measure imrpvoes, but if the model is really big
that can be time consuming if you do not want to keep
the weights and just find out what performance can a model get
with a set of hyperparameters, use this parameter to skip it,
but the model will not be loadable later on.
:type skip_save_model: Boolean
:param skip_save_progress: Disables saving
progress each epoch. By default Ludwig saves weights and stats
after each epoch for enabling resuming of training, but if
the model is really big that can be time consuming and will uses
twice as much space, use this parameter to skip it, but training
cannot be resumed later on.
:type skip_save_progress: Boolean
:param skip_save_log: Disables saving TensorBoard
logs. By default Ludwig saves logs for the TensorBoard, but if it
is not needed turning it off can slightly increase the
overall speed..
:type skip_save_log: Boolean
:param gpus: List of GPUs that are available for training.
:type gpus: List
:param gpu_fraction: Fraction of the memory of each GPU to use at
the beginning of the training. The memory may grow elastically.
:type gpu_fraction: Integer
:param random_seed: Random seed used for weights initialization,
splits and any other random function.
:type random_seed: Integer
:param debug: If true turns on tfdbg with inf_or_nan checks.
:type debug: Boolean
:returns: None
"""
if model_load_path is not None:
# Load model
if is_on_master():
print_boxed('LOADING MODEL')
logger.info('Loading model: {}\n'.format(model_load_path))
model, _ = load_model_and_definition(model_load_path)
else:
# Build model
if is_on_master():
print_boxed('BUILDING MODEL', print_fun=logger.debug)
model = Model(model_definition['input_features'],
model_definition['output_features'],
model_definition['combiner'],
model_definition['training'],
model_definition['preprocessing'],
use_horovod=use_horovod,
random_seed=random_seed,
debug=debug)
contrib_command("train_model", model, model_definition, model_load_path)
# Train model
if is_on_master():
print_boxed('TRAINING')
return model, model.train(training_set,
validation_set=validation_set,
test_set=test_set,
save_path=save_path,
resume=resume,
skip_save_model=skip_save_model,
skip_save_progress=skip_save_progress,
skip_save_log=skip_save_log,
gpus=gpus,
gpu_fraction=gpu_fraction,
random_seed=random_seed,
**model_definition['training'])
def train(training_set,
validation_set,
test_set,
model_definition,
save_path='model',
model_load_path=None,
resume=False,
skip_save_progress_weights=False,
gpus=None,
gpu_fraction=1.0,
use_horovod=False,
random_seed=default_random_seed,
debug=False):
"""
:param training_set: Dataset contaning training data
:type training_set: Dataset
:param validation_set: Dataset contaning validation data
:type validation_set: Datasetk
:param test_set: Dataset contaning test data.
:type test_set: Dataset
:param model_definition: Model definition which defines the different
parameters of the model, features, preprocessing and training.
:type model_definition: Dictionary
:param save_path: The path to save the model to.
:type save_path: filepath (str)
:param model_load_path: If this is specified the loaded model will be used
as initialization (useful for transfer learning).
:type model_load_path: filepath (str)
:param skip_save_progress_weights: Skips saving the weights at the end of
each epoch. If this is true, training cannot be resumed from the
exactly the state at the end of the previous epoch.
:type skip_save_progress_weights: Boolean
:param gpus: List of GPUs that are available for training.
:type gpus: List
:param gpu_fraction: Fraction of the memory of each GPU to use at
the beginning of the training. The memory may grow elastically.
:type gpu_fraction: Integer
:param random_seed: Random seed used for weights initialization,
splits and any other random function.
:type random_seed: Integer
:param debug: If true turns on tfdbg with inf_or_nan checks.
:type debug: Boolean
:returns: None
"""
if model_load_path is not None:
# Load model
if is_on_master():
print_boxed('LOADING MODEL')
logging.info('Loading model: {}\n'.format(model_load_path))
model, _ = load_model_and_definition(model_load_path)
else:
# Build model
if is_on_master():
print_boxed('BUILDING MODEL', print_fun=logging.debug)
model = Model(model_definition['input_features'],
model_definition['output_features'],
model_definition['combiner'],
model_definition['training'],
model_definition['preprocessing'],
use_horovod=use_horovod,
random_seed=random_seed,
debug=debug)
# Train model
if is_on_master():
print_boxed('TRAINING')
return model, model.train(
training_set,
validation_set=validation_set,
test_set=test_set,
save_path=save_path,
resume=resume,
skip_save_progress_weights=skip_save_progress_weights,
gpus=gpus,
gpu_fraction=gpu_fraction,
random_seed=random_seed,
**model_definition['training'])
