def run_movie(flags_obj):
"""Construct all necessary functions and call run_loop.
Args:
flags_obj: Object containing user specified flags.
"""
if flags_obj.download_if_missing:
movielens.download(dataset=flags_obj.dataset, data_dir=flags_obj.data_dir)
train_input_fn, eval_input_fn, model_column_fn = \
movielens_dataset.construct_input_fns(
dataset=flags_obj.dataset, data_dir=flags_obj.data_dir,
batch_size=flags_obj.batch_size, repeat=flags_obj.epochs_between_evals)
tensors_to_log = {
'loss': '{loss_prefix}head/weighted_loss/value'
}
wide_deep_run_loop.run_loop(
name="MovieLens", train_input_fn=train_input_fn,
eval_input_fn=eval_input_fn,
model_column_fn=model_column_fn,
build_estimator_fn=build_estimator,
flags_obj=flags_obj,
tensors_to_log=tensors_to_log,
early_stop=False)
def run_census(flags_obj):
"""Construct all necessary functions and call run_loop.
Args:
flags_obj: Object containing user specified flags.
"""
if flags_obj.download_if_missing:
census_dataset.download(flags_obj.data_dir)
train_file = os.path.join(flags_obj.data_dir, census_dataset.TRAINING_FILE)
test_file = os.path.join(flags_obj.data_dir, census_dataset.EVAL_FILE)
# Train and evaluate the model every `flags.epochs_between_evals` epochs.
def train_input_fn():
return census_dataset.input_fn(
train_file, flags_obj.epochs_between_evals, True, flags_obj.batch_size)
def eval_input_fn():
return census_dataset.input_fn(test_file, 1, False, flags_obj.batch_size)
tensors_to_log = {
'average_loss': '{loss_prefix}head/truediv',
'loss': '{loss_prefix}head/weighted_loss/Sum'
}
wide_deep_run_loop.run_loop(
name="Census Income", train_input_fn=train_input_fn,
eval_input_fn=eval_input_fn,
model_column_fn=census_dataset.build_model_columns,
build_estimator_fn=build_estimator,
flags_obj=flags_obj,
tensors_to_log=tensors_to_log,
early_stop=True)
def run_census(flags_obj):
"""Construct all necessary functions and call run_loop.
Args:
flags_obj: Object containing user specified flags.
"""
if flags_obj.download_if_missing:
census_dataset.download(flags_obj.data_dir)
train_file = os.path.join(flags_obj.data_dir, census_dataset.TRAINING_FILE)
test_file = os.path.join(flags_obj.data_dir, census_dataset.EVAL_FILE)
# Train and evaluate the model every `flags.epochs_between_evals` epochs.
def train_input_fn():
return census_dataset.input_fn(
train_file, flags_obj.epochs_between_evals, True, flags_obj.batch_size)
def eval_input_fn():
return census_dataset.input_fn(test_file, 1, False, flags_obj.batch_size)
tensors_to_log = {
'average_loss': '{loss_prefix}head/truediv',
'loss': '{loss_prefix}head/weighted_loss/Sum'
}
wide_deep_run_loop.run_loop(
name="Census Income", train_input_fn=train_input_fn,
eval_input_fn=eval_input_fn,
model_column_fn=census_dataset.build_model_columns,
build_estimator_fn=build_estimator,
flags_obj=flags_obj,
tensors_to_log=tensors_to_log,
early_stop=True)