def run(args):
assert args.mode in modes, f"Mode must in {modes}"
config = UserConfig(DEFAULT_YAML, args.config, learning=True)
speech_featurizer = SpeechFeaturizer(config["speech_config"])
text_featurizer = TextFeaturizer(config["decoder_config"])
if args.mode == "train":
tf.random.set_seed(2020)
if args.mixed_precision:
policy = tf.keras.mixed_precision.experimental.Policy("mixed_float16")
tf.keras.mixed_precision.experimental.set_policy(policy)
print("Enabled mixed precision training")
ctc_trainer = CTCTrainer(speech_featurizer, text_featurizer,
config["learning_config"]["running_config"],
args.mixed_precision)
if args.tfrecords:
train_dataset = ASRTFRecordDataset(
config["learning_config"]["dataset_config"]["train_paths"],
config["learning_config"]["dataset_config"]["tfrecords_dir"],
speech_featurizer, text_featurizer, "train",
augmentations=config["learning_config"]["augmentations"], shuffle=True,
)
eval_dataset = ASRTFRecordDataset(
config["learning_config"]["dataset_config"]["eval_paths"],
config["learning_config"]["dataset_config"]["tfrecords_dir"],
speech_featurizer, text_featurizer, "eval", shuffle=False
)
else:
train_dataset = ASRSliceDataset(
stage="train", speech_featurizer=speech_featurizer,
text_featurizer=text_featurizer,
data_paths=config["learning_config"]["dataset_config"]["train_paths"],
augmentations=config["learning_config"]["augmentations"], shuffle=True,
)
eval_dataset = ASRSliceDataset(
stage="eval", speech_featurizer=speech_featurizer,
text_featurizer=text_featurizer,
data_paths=config["learning_config"]["dataset_config"]["eval_paths"],
shuffle=False
)
# Build DS2 model
f, c = speech_featurizer.compute_feature_dim()
with ctc_trainer.strategy.scope():
satt_ds2_model = SelfAttentionDS2(input_shape=[None, f, c],
arch_config=config["model_config"],
num_classes=text_featurizer.num_classes)
satt_ds2_model._build([1, 50, f, c])
optimizer = create_optimizer(
name=config["learning_config"]["optimizer_config"]["name"],
d_model=config["model_config"]["att"]["head_size"],
**config["learning_config"]["optimizer_config"]["config"]
)
# Compile
ctc_trainer.compile(satt_ds2_model, optimizer, max_to_keep=args.max_ckpts)
ctc_trainer.fit(train_dataset, eval_dataset, args.eval_train_ratio)
if args.export:
if args.from_weights:
ctc_trainer.model.save_weights(args.export)
else:
ctc_trainer.model.save(args.export)
elif args.mode == "test":
tf.random.set_seed(0)
assert args.export
text_featurizer.add_scorer(
Scorer(**text_featurizer.decoder_config["lm_config"],
vocabulary=text_featurizer.vocab_array))
# Build DS2 model
f, c = speech_featurizer.compute_feature_dim()
satt_ds2_model = SelfAttentionDS2(input_shape=[None, f, c],
arch_config=config["model_config"],
num_classes=text_featurizer.num_classes)
satt_ds2_model._build([1, 50, f, c])
satt_ds2_model.summary(line_length=100)
optimizer = create_optimizer(
name=config["learning_config"]["optimizer_config"]["name"],
d_model=config["model_config"]["att"]["head_size"],
**config["learning_config"]["optimizer_config"]["config"]
)
batch_size = config["learning_config"]["running_config"]["batch_size"]
if args.tfrecords:
test_dataset = ASRTFRecordDataset(
config["learning_config"]["dataset_config"]["test_paths"],
config["learning_config"]["dataset_config"]["tfrecords_dir"],
speech_featurizer, text_featurizer, "test",
augmentations=config["learning_config"]["augmentations"], shuffle=False
).create(batch_size * args.eval_train_ratio)
else:
test_dataset = ASRSliceDataset(
stage="test", speech_featurizer=speech_featurizer,
text_featurizer=text_featurizer,
data_paths=config["learning_config"]["dataset_config"]["test_paths"],
augmentations=config["learning_config"]["augmentations"], shuffle=False
).create(batch_size * args.eval_train_ratio)
ctc_tester = BaseTester(
config=config["learning_config"]["running_config"],
saved_path=args.export, from_weights=args.from_weights
)
ctc_tester.compile(satt_ds2_model, speech_featurizer, text_featurizer)
ctc_tester.run(test_dataset)
else:
assert args.export
# Build DS2 model
f, c = speech_featurizer.compute_feature_dim()
satt_ds2_model = SelfAttentionDS2(input_shape=[None, f, c],
arch_config=config["model_config"],
num_classes=text_featurizer.num_classes)
satt_ds2_model._build([1, 50, f, c])
optimizer = create_optimizer(
name=config["learning_config"]["optimizer_config"]["name"],
d_model=config["model_config"]["att"]["head_size"],
**config["learning_config"]["optimizer_config"]["config"]
)
def save_func(**kwargs):
if args.from_weights:
kwargs["model"].save_weights(args.export)
else:
kwargs["model"].save(args.export)
save_from_checkpoint(func=save_func,
outdir=config["learning_config"]["running_config"]["outdir"],
model=satt_ds2_model, optimizer=optimizer)
speech_featurizer=speech_featurizer,
text_featurizer=text_featurizer,
data_paths=config["learning_config"]["dataset_config"]["train_paths"],
augmentations=config["learning_config"]["augmentations"],
stage="train",
cache=args.cache,
shuffle=True)
eval_dataset = ASRSliceDataset(
speech_featurizer=speech_featurizer,
text_featurizer=text_featurizer,
data_paths=config["learning_config"]["dataset_config"]["eval_paths"],
stage="eval",
cache=args.cache,
shuffle=True)
ctc_trainer = CTCTrainer(text_featurizer,
config["learning_config"]["running_config"])
# Build DS2 model
with ctc_trainer.strategy.scope():
ds2_model = DeepSpeech2(input_shape=speech_featurizer.shape,
arch_config=config["model_config"],
num_classes=text_featurizer.num_classes,
name="deepspeech2")
ds2_model._build(speech_featurizer.shape)
ds2_model.summary(line_length=150)
# Compile
ctc_trainer.compile(ds2_model,
config["learning_config"]["optimizer_config"],
max_to_keep=args.max_ckpts)
ctc_trainer.fit(train_dataset,
eval_dataset,
def main():
tf.keras.backend.clear_session()
parser = argparse.ArgumentParser(prog="Deep Speech 2 Training")
parser.add_argument("--config",
"-c",
type=str,
default=DEFAULT_YAML,
help="The file path of model configuration file")
parser.add_argument("--export",
"-e",
type=str,
default=None,
help="Path to the model file to be exported")
parser.add_argument("--mixed_precision",
type=bool,
default=False,
help="Whether to use mixed precision training")
parser.add_argument("--save_weights",
type=bool,
default=False,
help="Whether to save or load only weights")
parser.add_argument("--max_ckpts",
type=int,
default=10,
help="Max number of checkpoints to keep")
parser.add_argument(
"--eval_train_ratio",
type=int,
default=1,
help="ratio between train batch size and eval batch size")
parser.add_argument("--tfrecords",
type=bool,
default=False,
help="Whether to use tfrecords dataset")
args = parser.parse_args()
config = UserConfig(DEFAULT_YAML, args.config, learning=True)
speech_featurizer = TFSpeechFeaturizer(config["speech_config"])
text_featurizer = TextFeaturizer(config["decoder_config"])
tf.random.set_seed(2020)
if args.mixed_precision:
policy = tf.keras.mixed_precision.experimental.Policy("mixed_float16")
tf.keras.mixed_precision.experimental.set_policy(policy)
print("Enabled mixed precision training")
if args.tfrecords:
train_dataset = ASRTFRecordDataset(
config["learning_config"]["dataset_config"]["train_paths"],
config["learning_config"]["dataset_config"]["tfrecords_dir"],
speech_featurizer,
text_featurizer,
"train",
augmentations=config["learning_config"]["augmentations"],
shuffle=True,
)
eval_dataset = ASRTFRecordDataset(
config["learning_config"]["dataset_config"]["eval_paths"],
config["learning_config"]["dataset_config"]["tfrecords_dir"],
speech_featurizer,
text_featurizer,
"eval",
shuffle=False)
else:
train_dataset = ASRSliceDataset(
stage="train",
speech_featurizer=speech_featurizer,
text_featurizer=text_featurizer,
data_paths=config["learning_config"]["dataset_config"]
["eval_paths"],
augmentations=config["learning_config"]["augmentations"],
shuffle=True)
eval_dataset = ASRSliceDataset(stage="train",
speech_featurizer=speech_featurizer,
text_featurizer=text_featurizer,
data_paths=config["learning_config"]
["dataset_config"]["eval_paths"],
shuffle=True)
ctc_trainer = CTCTrainer(speech_featurizer, text_featurizer,
config["learning_config"]["running_config"],
args.mixed_precision)
# Build DS2 model
f, c = speech_featurizer.compute_feature_dim()
with ctc_trainer.strategy.scope():
ds2_model = DeepSpeech2(input_shape=[None, f, c],
arch_config=config["model_config"],
num_classes=text_featurizer.num_classes,
name="deepspeech2")
ds2_model._build([1, 50, f, c])
# Compile
ctc_trainer.compile(ds2_model,
config["learning_config"]["optimizer_config"],
max_to_keep=args.max_ckpts)
ctc_trainer.fit(train_dataset, eval_dataset, args.eval_train_ratio)
if args.export:
if args.save_weights:
ctc_trainer.model.save_weights(args.export)
else:
ctc_trainer.model.save(args.export)