Medix-AI/coqui-tts
Medix-AI
/
coqui-tts
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

bug fix on train_encoder and unit tests

This commit is contained in:
Edresson 2021-07-11 12:00:39 -03:00
parent 8fbadad68e
commit 4eac1c4651
2 changed files with 31 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
TTS/bin/train_encoder.py
View File

@ -164,7 +164,7 @@ def main(args): # pylint: disable=redefined-outer-name
elif c.loss == "angleproto":
criterion = AngleProtoLoss()
elif c.loss == "softmaxproto":
criterion = SoftmaxAngleProtoLoss(c.model["proj_dim"], num_speakers)
criterion = SoftmaxAngleProtoLoss(c.model_params["proj_dim"], num_speakers)
else:
raise Exception("The %s not is a loss supported" % c.loss)

49
tests/test_speaker_encoder_train.py
View File

@ -6,7 +6,18 @@ from tests import get_device_id, get_tests_output_path, run_cli
from TTS.config.shared_configs import BaseAudioConfig
from TTS.speaker_encoder.speaker_encoder_config import SpeakerEncoderConfig
config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
def run_test_train():
command = (
f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_encoder.py --config_path {config_path} "
f"--coqpit.output_path {output_path} "
"--coqpit.datasets.0.name ljspeech "
"--coqpit.datasets.0.meta_file_train metadata.csv "
"--coqpit.datasets.0.meta_file_val metadata.csv "
"--coqpit.datasets.0.path tests/data/ljspeech "
)
run_cli(command)
config_path = os.path.join(get_tests_output_path(), "test_speaker_encoder_config.json")
output_path = os.path.join(get_tests_output_path(), "train_outputs")
config = SpeakerEncoderConfig(
@ -24,16 +35,9 @@ config.audio.do_trim_silence = True
config.audio.trim_db = 60
config.save_json(config_path)
print(config)
# train the model for one epoch
command_train = (
f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_encoder.py --config_path {config_path} "
f"--coqpit.output_path {output_path} "
"--coqpit.datasets.0.name ljspeech "
"--coqpit.datasets.0.meta_file_train metadata.csv "
"--coqpit.datasets.0.meta_file_val metadata.csv "
"--coqpit.datasets.0.path tests/data/ljspeech "
)
run_cli(command_train)
run_test_train()
# Find latest folder
continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
@ -50,15 +54,7 @@ config.model_params["model_name"] = "resnet"
config.save_json(config_path)
# train the model for one epoch
command_train = (
f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_encoder.py --config_path {config_path} "
f"--coqpit.output_path {output_path} "
"--coqpit.datasets.0.name ljspeech "
"--coqpit.datasets.0.meta_file_train metadata.csv "
"--coqpit.datasets.0.meta_file_val metadata.csv "
"--coqpit.datasets.0.path tests/data/ljspeech "
)
run_cli(command_train)
run_test_train()
# Find latest folder
continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
@ -69,3 +65,18 @@ command_train = (
)
run_cli(command_train)
shutil.rmtree(continue_path)
# test model with ge2e loss function
config.loss = "ge2e"
config.save_json(config_path)
run_test_train()
# test model with angleproto loss function
config.loss = "angleproto"
config.save_json(config_path)
run_test_train()
# test model with softmaxproto loss function
config.loss = "softmaxproto"
config.save_json(config_path)
run_test_train()