def _do_evaluation(model: Module,
settings_data: MutableMapping[str, Any],
settings_io: MutableMapping[str, Any],
indices_list: MutableSequence[str]) \
-> None:
"""Evaluation of an optimized model.
:param model: Model to use.
:type model: torch.nn.Module
:param settings_data: Data settings to use.
:type settings_data: dict
:param indices_list: Sequence with the words of the captions.
:type indices_list: list[str]
"""
model.eval()
logger_main = logger.bind(is_caption=False, indent=1)
data_path_evaluation = Path(
settings_io['root_dirs']['data'],
settings_io['dataset']['features_dirs']['output'],
settings_io['dataset']['features_dirs']['evaluation'])
logger_main.info('Getting evaluation data')
validation_data = get_clotho_loader(
settings_io['dataset']['features_dirs']['evaluation'],
is_training=False,
settings_data=settings_data,
settings_io=settings_io)
logger_main.info('Done')
text_sep = '-' * 100
starting_text = 'Starting evaluation on evaluation data'
logger_main.info(starting_text)
logger.bind(is_caption=True,
indent=0).info(f'{text_sep}\n{text_sep}\n{text_sep}\n\n')
logger.bind(is_caption=True, indent=0).info(f'{starting_text}.\n\n')
with no_grad():
evaluation_outputs = module_epoch_passing(data=validation_data,
module=model,
objective=None,
optimizer=None,
epoch=-1,
max_epoch=-1)
captions_pred, captions_gt = _decode_outputs(
evaluation_outputs[1],
evaluation_outputs[2],
indices_object=indices_list,
file_names=sorted(list(data_path_evaluation.iterdir())),
eos_token='<eos>',
print_to_console=False)
logger_main.info('Evaluation done')
metrics = evaluate_metrics(captions_pred, captions_gt)
for metric, values in metrics.items():
logger_main.info(f'{metric:<7s}: {values["score"]:7.4f}')
def _do_training(model: Module,
settings_training: MutableMapping[
str, Union[Any, MutableMapping[str, Any]]],
settings_data: MutableMapping[
str, Union[Any, MutableMapping[str, Any]]],
settings_io: MutableMapping[
str, Union[Any, MutableMapping[str, Any]]],
model_file_name: str,
model_dir: Path,
indices_list: MutableSequence[str]) \
-> None:
"""Optimization of the model.
:param model: Model to optimize.
:type model: torch.nn.Module
:param settings_training: Training settings to use.
:type settings_training: dict
:param settings_data: Training data settings to use.
:type settings_data: dict
:param settings_io: Data I/O settings to use.
:type settings_io: dict
:param model_file_name: File name of the model.
:type model_file_name: str
:param model_dir: Directory to serialize the model to.
:type model_dir: pathlib.Path
:param indices_list: A sequence with the words.
:type indices_list: list[str]
"""
# Initialize variables for the training process
prv_training_loss = 1e8
patience: int = settings_training['patience']
loss_thr: float = settings_training['loss_thr']
patience_counter = 0
best_epoch = 0
# Initialize logger
logger_main = logger.bind(is_caption=False, indent=1)
# Inform that we start getting the data
logger_main.info('Getting training data')
# Get training data and count the amount of batches
training_data = get_clotho_loader(
settings_io['dataset']['features_dirs']['development'],
is_training=True,
settings_data=settings_data,
settings_io=settings_io)
logger_main.info('Done')
# Initialize loss and optimizer objects
objective = CrossEntropyLoss()
optimizer = Adam(params=model.parameters(),
lr=settings_training['optimizer']['lr'])
# Inform that we start training
logger_main.info('Starting training')
model.train()
for epoch in range(settings_training['nb_epochs']):
# Log starting time
start_time = time()
# Do a complete pass over our training data
epoch_output = module_epoch_passing(
data=training_data,
module=model,
objective=objective,
optimizer=optimizer,
grad_norm=settings_training['grad_norm']['norm'],
grad_norm_val=settings_training['grad_norm']['value'])
objective_output, output_y_hat, output_y, f_names = epoch_output
# Get mean loss of training and print it with logger
training_loss = objective_output.mean().item()
logger_main.info(f'Epoch: {epoch:05d} -- '
f'Training loss: {training_loss:>7.4f} | '
f'Time: {time() - start_time:>5.3f}')
# Check if we have to decode captions for the current epoch
if divmod(epoch + 1,
settings_training['text_output_every_nb_epochs'])[-1] == 0:
# Get the subset of files for decoding their captions
sampling_indices = sorted(
randperm(len(output_y_hat))
[:settings_training['nb_examples_to_sample']].tolist())
# Do the decoding
_decode_outputs(*zip(*[[output_y_hat[i], output_y[i]]
for i in sampling_indices]),
indices_object=indices_list,
file_names=[
Path(f_names[i_f_name])
for i_f_name in sampling_indices
],
eos_token='<eos>',
print_to_console=False)
# Check improvement of loss
if prv_training_loss - training_loss > loss_thr:
# Log the current loss
prv_training_loss = training_loss
# Log the current epoch
best_epoch = epoch
# Serialize the model keeping the epoch
pt_save(
model.state_dict(),
str(
model_dir.joinpath(
f'epoch_{best_epoch:05d}_{model_file_name}')))
# Zero out the patience
patience_counter = 0
else:
# Increase patience counter
patience_counter += 1
# Serialize the model and optimizer.
for pt_obj, save_str in zip([model, optimizer], ['', '_optimizer']):
pt_save(
pt_obj.state_dict(),
str(model_dir.joinpath(f'latest{save_str}_{model_file_name}')))
# Check for stopping criteria
if patience_counter >= patience:
logger_main.info('No lower training loss for '
f'{patience_counter} epochs. '
'Training stops.')
break
# Inform that we are done
logger_main.info('Training done')
# Load best model
model.load_state_dict(
pt_load(
str(model_dir.joinpath(
f'epoch_{best_epoch:05d}_{model_file_name}'))))
def _do_testing(model: Module,
settings_data: MutableMapping[str, Any],
settings_io: MutableMapping[str, Any],
indices_list: MutableSequence[str]) \
-> None:
"""Evaluation of an optimized model.
:param model: Model to use.
:type model: torch.nn.Module
:param settings_data: Data settings to use.
:type settings_data: dict
:param indices_list: Sequence with the words of the captions.
:type indices_list: list[str]
"""
model.eval()
logger_main = logger.bind(is_caption=False, indent=1)
data_path_test = Path(settings_io['root_dirs']['data'],
settings_io['dataset']['features_dirs']['output'],
settings_io['dataset']['features_dirs']['test'])
logger_main.info('Getting test data')
test_data = get_clotho_loader(
settings_io['dataset']['features_dirs']['test'],
is_training=False,
settings_data=settings_data,
settings_io=settings_io)
logger_main.info('Done')
text_sep = '-' * 100
starting_text = 'Starting testing on test data'
logger_main.info(starting_text)
logger.bind(is_caption=True,
indent=0).info(f'{text_sep}\n{text_sep}\n{text_sep}\n\n')
logger.bind(is_caption=True, indent=0).info(f'{starting_text}.\n\n')
with no_grad():
test_outputs = module_epoch_passing(data=test_data,
module=model,
objective=None,
optimizer=None)
captions_pred, _ = _decode_outputs(test_outputs[1],
test_outputs[2],
indices_object=indices_list,
file_names=list(
data_path_test.iterdir()),
eos_token='<eos>',
print_to_console=False)
# clotho_file_{file_name} to {file_name}.wav
for i, entry in enumerate(captions_pred):
entry['file_name'] = entry['file_name']\
.replace('clotho_file_', '') + '.wav'
captions_pred[i] = entry
submission_dir = Path().joinpath(
settings_io['root_dirs']['outputs'],
settings_io['submissions']['submissions_dir'])
submission_dir.mkdir(parents=True, exist_ok=True)
csv_functions.write_csv_file(captions_pred,
settings_io['submissions']['submission_file'],
submission_dir,
add_timestamp=True)
logger_main.info('Testing done')
settings_training = settings['dnn_training_settings']['training'],
settings_data = settings['dnn_training_settings']['data'],
settings_io = settings['dirs_and_files']
indices_list = _load_indices_file(
settings['dirs_and_files'], settings['dnn_training_settings']['data'])
data_path_evaluation = Path(
settings_io['root_dirs']['data'],
settings_io['dataset']['features_dirs']['output'],
settings_io['dataset']['features_dirs']['evaluation'])
validation_data = get_clotho_loader(
settings_io['dataset']['features_dirs']['evaluation'],
is_training=False,
settings_data=settings['dnn_training_settings']['data'],
settings_io=settings['dirs_and_files'])
print("load settings end")
print("load pre-trained models start")
pretrain_emb = align_word_embedding(r'data/pickles/words_list.p',
r'outputs/models/w2v_192.mod', ntoken,
nhid)
# pretrain_cnn = torch.load(Path('outputs/models/TagModel_60.pt'), map_location='cuda')
model = TransformerModel(ntoken,
ninp,
nhead,
nhid,
print("load settings start")
settings = load_yaml_file(Path(file_dir, f'{config_file}.{file_ext}'))
settings_training = settings['dnn_training_settings']['training'],
settings_data = settings['dnn_training_settings']['data'],
settings_io = settings['dirs_and_files']
indices_list = _load_indices_file(
settings['dirs_and_files'], settings['dnn_training_settings']['data'])
train_data = get_clotho_loader(split='development',
data_dir=param.data_path_development,
is_training=True,
input_field_name=param.input_field_name,
output_field_name=param.output_field_name,
batch_size=param.batch_size,
augment=True)
eval_data = get_clotho_loader(split='evaluation',
data_dir=param.data_path_evaluation,
is_training=False,
input_field_name=param.input_field_name,
output_field_name=param.output_field_name,
batch_size=param.batch_size)
print("load settings end")
print("load pre-trained models start")
pretrain_emb = align_word_embedding(r'data/pickles/words_list.p',
