def main():
args = parser.parse_args()
##################################################
# DATSET
##################################################
if args.model_save_path is not None:
# Load a config file (.yml)
params = load_config(args.config_path)
# NOTE: Retrain the saved model from the last checkpoint
elif args.saved_model_path is not None:
params = load_config(os.path.join(args.saved_model_path, 'config.yml'))
else:
raise ValueError("Set model_save_path or saved_model_path.")
# Load dataset
train_data = Dataset(data_save_path=args.data_save_path,
backend=params['backend'],
input_freq=params['input_freq'],
use_delta=params['use_delta'],
use_double_delta=params['use_double_delta'],
data_type='train',
data_size=params['data_size'],
label_type=params['label_type'],
batch_size=params['batch_size'],
max_epoch=params['num_epoch'],
splice=params['splice'],
num_stack=params['num_stack'],
num_skip=params['num_skip'],
min_frame_num=params['min_frame_num'],
sort_utt=True,
sort_stop_epoch=params['sort_stop_epoch'],
tool=params['tool'],
num_enque=None,
dynamic_batching=params['dynamic_batching'])
dev_data = Dataset(data_save_path=args.data_save_path,
backend=params['backend'],
input_freq=params['input_freq'],
use_delta=params['use_delta'],
use_double_delta=params['use_double_delta'],
data_type='dev',
data_size=params['data_size'],
label_type=params['label_type'],
batch_size=params['batch_size'],
splice=params['splice'],
num_stack=params['num_stack'],
num_skip=params['num_skip'],
shuffle=True,
tool=params['tool'])
eval1_data = Dataset(data_save_path=args.data_save_path,
backend=params['backend'],
input_freq=params['input_freq'],
use_delta=params['use_delta'],
use_double_delta=params['use_double_delta'],
data_type='eval1',
data_size=params['data_size'],
label_type=params['label_type'],
batch_size=params['batch_size'],
splice=params['splice'],
num_stack=params['num_stack'],
num_skip=params['num_skip'],
tool=params['tool'])
params['num_classes'] = train_data.num_classes
params['num_classes_sub'] = train_data.num_classes
##################################################
# MODEL
##################################################
# Model setting
model = load(model_type=params['model_type'],
params=params,
backend=params['backend'])
if args.model_save_path is not None:
# Set save path
save_path = mkdir_join(args.model_save_path, params['backend'],
params['model_type'], params['label_type'],
params['data_size'], model.name)
model.set_save_path(save_path)
# Save config file
save_config(config_path=args.config_path, save_path=model.save_path)
# Setting for logging
logger = set_logger(model.save_path)
if os.path.isdir(params['char_init']):
# NOTE: Start training from the pre-trained character model
model.load_checkpoint(save_path=params['char_init'],
epoch=-1,
load_pretrained_model=True)
# Count total parameters
for name in sorted(list(model.num_params_dict.keys())):
num_params = model.num_params_dict[name]
logger.info("%s %d" % (name, num_params))
logger.info("Total %.3f M parameters" %
(model.total_parameters / 1000000))
# Define optimizer
model.set_optimizer(optimizer=params['optimizer'],
learning_rate_init=float(params['learning_rate']),
weight_decay=float(params['weight_decay']),
clip_grad_norm=params['clip_grad_norm'],
lr_schedule=False,
factor=params['decay_rate'],
patience_epoch=params['decay_patient_epoch'])
epoch, step = 1, 0
learning_rate = float(params['learning_rate'])
metric_dev_best = 1
# NOTE: Retrain the saved model from the last checkpoint
elif args.saved_model_path is not None:
# Set save path
model.save_path = args.saved_model_path
# Setting for logging
logger = set_logger(model.save_path, restart=True)
# Define optimizer
model.set_optimizer(
optimizer=params['optimizer'],
learning_rate_init=float(params['learning_rate']), # on-the-fly
weight_decay=float(params['weight_decay']),
clip_grad_norm=params['clip_grad_norm'],
lr_schedule=False,
factor=params['decay_rate'],
patience_epoch=params['decay_patient_epoch'])
# Restore the last saved model
epoch, step, learning_rate, metric_dev_best = model.load_checkpoint(
save_path=args.saved_model_path, epoch=-1, restart=True)
else:
raise ValueError("Set model_save_path or saved_model_path.")
train_data.epoch = epoch - 1
# GPU setting
model.set_cuda(deterministic=False, benchmark=True)
logger.info('PID: %s' % os.getpid())
logger.info('USERNAME: %s' % os.uname()[1])
# Set process name
setproctitle('csj_' + params['backend'] + '_' + params['model_type'] +
'_' + params['label_type'] + '_' + params['data_size'])
##################################################
# TRAINING LOOP
##################################################
# Define learning rate controller
lr_controller = Controller(
learning_rate_init=learning_rate,
backend=params['backend'],
decay_start_epoch=params['decay_start_epoch'],
decay_rate=params['decay_rate'],
decay_patient_epoch=params['decay_patient_epoch'],
lower_better=True)
# Setting for tensorboard
if params['backend'] == 'pytorch':
tf_writer = SummaryWriter(model.save_path)
# Train model
csv_steps, csv_loss_train, csv_loss_dev = [], [], []
start_time_train = time.time()
start_time_epoch = time.time()
start_time_step = time.time()
not_improved_epoch = 0
best_model = model
loss_train_mean = 0.
pbar_epoch = tqdm(total=len(train_data))
while True:
# Compute loss in the training set (including parameter update)
batch_train, is_new_epoch = train_data.next()
model, loss_train = train_step(model, batch_train,
params['clip_grad_norm'],
params['backend'])
loss_train_mean += loss_train
pbar_epoch.update(len(batch_train['xs']))
if (step + 1) % params['print_step'] == 0:
# Compute loss in the dev set
batch_dev = dev_data.next()[0]
loss_dev = model(batch_dev['xs'],
batch_dev['ys'],
batch_dev['x_lens'],
batch_dev['y_lens'],
is_eval=True)
loss_train_mean /= params['print_step']
csv_steps.append(step)
csv_loss_train.append(loss_train_mean)
csv_loss_dev.append(loss_dev)
# Logging by tensorboard
if params['backend'] == 'pytorch':
tf_writer.add_scalar('train/loss', loss_train_mean, step + 1)
tf_writer.add_scalar('dev/loss', loss_dev, step + 1)
# for name, param in model.named_parameters():
# name = name.replace('.', '/')
# tf_writer.add_histogram(
# name, param.data.cpu().numpy(), step + 1)
# tf_writer.add_histogram(
# name + '/grad', param.grad.data.cpu().numpy(), step + 1)
# TODO: fix this
duration_step = time.time() - start_time_step
logger.info(
"...Step:%d(epoch:%.3f) loss:%.3f(%.3f)/lr:%.5f/batch:%d/x_lens:%d (%.3f min)"
% (step + 1, train_data.epoch_detail, loss_train_mean,
loss_dev, learning_rate, train_data.current_batch_size,
max(batch_train['x_lens']) * params['num_stack'],
duration_step / 60))
start_time_step = time.time()
loss_train_mean = 0.
step += 1
# Save checkpoint and evaluate model per epoch
if is_new_epoch:
duration_epoch = time.time() - start_time_epoch
logger.info('===== EPOCH:%d (%.3f min) =====' %
(epoch, duration_epoch / 60))
# Save fugure of loss
plot_loss(csv_loss_train,
csv_loss_dev,
csv_steps,
save_path=model.save_path)
if epoch < params['eval_start_epoch']:
# Save the model
model.save_checkpoint(model.save_path, epoch, step,
learning_rate, metric_dev_best)
else:
start_time_eval = time.time()
# dev
if params['label_type'] == 'word':
metric_dev, _ = eval_word(
models=[model],
dataset=dev_data,
eval_batch_size=1,
beam_width=1,
max_decode_len=MAX_DECODE_LEN_WORD)
logger.info(' WER (dev): %.3f %%' % (metric_dev * 100))
else:
wer_dev, metric_dev, _ = eval_char(
models=[model],
dataset=dev_data,
eval_batch_size=1,
beam_width=1,
max_decode_len=MAX_DECODE_LEN_CHAR)
logger.info(' WER / CER (dev): %.3f / %.3f %%' %
((wer_dev * 100), (metric_dev * 100)))
if metric_dev < metric_dev_best:
metric_dev_best = metric_dev
not_improved_epoch = 0
best_model = copy.deepcopy(model)
logger.info('||||| Best Score |||||')
# Save the model
model.save_checkpoint(model.save_path, epoch, step,
learning_rate, metric_dev_best)
# test
if params['label_type'] == 'word':
wer_eval1, _ = eval_word(
models=[model],
dataset=eval1_data,
eval_batch_size=1,
beam_width=1,
max_decode_len=MAX_DECODE_LEN_CHAR)
logger.info(' WER (eval1): %.3f %%' %
(wer_eval1 * 100))
else:
wer_eval1, cer_eval1, _ = eval_char(
models=[model],
dataset=eval1_data,
eval_batch_size=1,
beam_width=1,
max_decode_len=MAX_DECODE_LEN_CHAR)
logger.info(' WER / CER (eval1): %.3f / %.3f %%' %
((wer_eval1 * 100), (cer_eval1 * 100)))
else:
not_improved_epoch += 1
duration_eval = time.time() - start_time_eval
logger.info('Evaluation time: %.3f min' % (duration_eval / 60))
# Early stopping
if not_improved_epoch == params['not_improved_patient_epoch']:
break
# Update learning rate
model.optimizer, learning_rate = lr_controller.decay_lr(
optimizer=model.optimizer,
learning_rate=learning_rate,
epoch=epoch,
value=metric_dev)
if epoch == params['convert_to_sgd_epoch']:
# Convert to fine-tuning stage
model.set_optimizer(
'sgd',
learning_rate_init=learning_rate,
weight_decay=float(params['weight_decay']),
clip_grad_norm=params['clip_grad_norm'],
lr_schedule=False,
factor=params['decay_rate'],
patience_epoch=params['decay_patient_epoch'])
logger.info('========== Convert to SGD ==========')
# Inject Gaussian noise to all parameters
if float(params['weight_noise_std']) > 0:
model.weight_noise_injection = True
pbar_epoch = tqdm(total=len(train_data))
print('========== EPOCH:%d (%.3f min) ==========' %
(epoch, duration_epoch / 60))
if epoch == params['num_epoch']:
break
start_time_step = time.time()
start_time_epoch = time.time()
epoch += 1
# TODO: evaluate the best model by beam search here
duration_train = time.time() - start_time_train
logger.info('Total time: %.3f hour' % (duration_train / 3600))
if params['backend'] == 'pytorch':
tf_writer.close()
# Training was finished correctly
with open(os.path.join(model.save_path, 'COMPLETE'), 'w') as f:
f.write('')
def main():
args = parser.parse_args()
# Load a config file (.yml)
params = load_config(join(args.model_path, 'config.yml'), is_eval=True)
wer_mean, cer_mean = 0, 0
with open(join(args.model_path, 'RESULTS'), 'w') as f:
for i, data_type in enumerate(['eval1', 'eval2', 'eval3']):
# Load dataset
eval_data = Dataset(
data_save_path=args.data_save_path,
backend=params['backend'],
input_freq=params['input_freq'],
use_delta=params['use_delta'],
use_double_delta=params['use_double_delta'],
data_type=data_type, data_size=params['data_size'],
label_type=params['label_type'],
batch_size=args.eval_batch_size, splice=params['splice'],
num_stack=params['num_stack'], num_skip=params['num_skip'],
shuffle=False, tool=params['tool'])
if i == 0:
params['num_classes'] = eval_data.num_classes
# Load model
model = load(model_type=params['model_type'],
params=params,
backend=params['backend'])
# Restore the saved parameters
model.load_checkpoint(
save_path=args.model_path, epoch=args.epoch)
# GPU setting
model.set_cuda(deterministic=False, benchmark=True)
print('beam width: %d' % args.beam_width)
f.write('beam width: %d\n' % args.beam_width)
if params['label_type'] == 'word':
wer, df = eval_word(
models=[model],
dataset=eval_data,
eval_batch_size=args.eval_batch_size,
beam_width=args.beam_width,
max_decode_len=MAX_DECODE_LEN_WORD,
length_penalty=args.length_penalty,
progressbar=True)
wer_mean += wer
print(' WER (%s): %.3f %%' % (data_type, (wer * 100)))
f.write(' WER (%s): %.3f %%' % (data_type, (wer * 100)))
print(df)
else:
wer, cer, df = eval_char(
models=[model],
dataset=eval_data,
eval_batch_size=args.eval_batch_size,
beam_width=args.beam_width,
max_decode_len=MAX_DECODE_LEN_CHAR,
length_penalty=args.length_penalty,
progressbar=True)
wer_mean += wer
cer_mean += cer
print(' WER / CER (%s, sub): %.3f / %.3f %%' %
(data_type, (wer * 100), (cer * 100)))
f.write(' WER / CER (%s, sub): %.3f / %.3f %%' %
(data_type, (wer * 100), (cer * 100)))
print(df)
if params['label_type'] == 'word':
print(' WER (mean): %.3f %%' % (wer * 100 / 3))
f.write(' WER (mean): %.3f %%' % (wer * 100 / 3))
else:
print(' WER / CER (mean): %.3f / %.3f %%' %
((wer * 100 / 3), (cer * 100 / 3)))
f.write(' WER / CER (mean): %.3f / %.3f %%' %
((wer * 100 / 3), (cer * 100 / 3)))
def main():
args = parser.parse_args()
# Load a config file (.yml)
params = load_config(join(args.model_path, 'config.yml'), is_eval=True)
# Setting for logging
logger = set_logger(args.model_path)
wer_mean, wer_sub_mean, cer_sub_mean = 0, 0, 0
for i, data_type in enumerate(['eval1', 'eval2', 'eval3']):
# Load dataset
dataset = Dataset(data_save_path=args.data_save_path,
backend=params['backend'],
input_freq=params['input_freq'],
use_delta=params['use_delta'],
use_double_delta=params['use_double_delta'],
data_type=data_type,
data_size=params['data_size'],
label_type=params['label_type'],
label_type_sub=params['label_type_sub'],
batch_size=args.eval_batch_size,
splice=params['splice'],
num_stack=params['num_stack'],
num_skip=params['num_skip'],
shuffle=False,
tool=params['tool'])
if i == 0:
params['num_classes'] = dataset.num_classes
params['num_classes_sub'] = dataset.num_classes_sub
# Load model
model = load(model_type=params['model_type'],
params=params,
backend=params['backend'])
# Restore the saved parameters
epoch, _, _, _ = model.load_checkpoint(save_path=args.model_path,
epoch=args.epoch)
# GPU setting
model.set_cuda(deterministic=False, benchmark=True)
logger.info('beam width (main): %d\n' % args.beam_width)
logger.info('beam width (sub) : %d\n' % args.beam_width_sub)
logger.info('epoch: %d' % (epoch - 1))
logger.info('a2c oracle: %s\n' % str(args.a2c_oracle))
logger.info('resolving_unk: %s\n' % str(args.resolving_unk))
logger.info('joint_decoding: %s\n' % str(args.joint_decoding))
logger.info('score_sub_weight : %f' % args.score_sub_weight)
wer, df = eval_word(models=[model],
dataset=dataset,
eval_batch_size=args.eval_batch_size,
beam_width=args.beam_width,
max_decode_len=MAX_DECODE_LEN_WORD,
min_decode_len=MIN_DECODE_LEN_WORD,
beam_width_sub=args.beam_width_sub,
max_decode_len_sub=MAX_DECODE_LEN_CHAR,
min_decode_len_sub=MIN_DECODE_LEN_CHAR,
length_penalty=args.length_penalty,
coverage_penalty=args.coverage_penalty,
progressbar=True,
resolving_unk=args.resolving_unk,
a2c_oracle=args.a2c_oracle,
joint_decoding=args.joint_decoding,
score_sub_weight=args.score_sub_weight)
wer_mean += wer
logger.info(' WER (%s, main): %.3f %%' % (data_type, (wer * 100)))
logger.info(df)
wer_sub, cer_sub, df_sub = eval_char(
models=[model],
dataset=dataset,
eval_batch_size=args.eval_batch_size,
beam_width=args.beam_width_sub,
max_decode_len=MAX_DECODE_LEN_CHAR,
min_decode_len=MIN_DECODE_LEN_CHAR,
length_penalty=args.length_penalty,
coverage_penalty=args.coverage_penalty,
progressbar=True)
wer_sub_mean += wer_sub
cer_sub_mean += cer_sub
logger.info(' WER / CER (%s, sub): %.3f / %.3f %%' % (data_type,
(wer_sub * 100),
(cer_sub * 100)))
logger.info(df_sub)
logger.info(' WER (mean, main): %.3f %%' % (wer_mean * 100 / 3))
logger.info(' WER / CER (mean, sub): %.3f / %.3f %%' %
((wer_sub_mean * 100 / 3), (cer_sub_mean * 100 / 3)))