|
|
@ -59,6 +59,10 @@ model_alias = {
|
|
|
|
"paddlespeech.t2s.models.hifigan:HiFiGANGenerator",
|
|
|
|
"paddlespeech.t2s.models.hifigan:HiFiGANGenerator",
|
|
|
|
"hifigan_inference":
|
|
|
|
"hifigan_inference":
|
|
|
|
"paddlespeech.t2s.models.hifigan:HiFiGANInference",
|
|
|
|
"paddlespeech.t2s.models.hifigan:HiFiGANInference",
|
|
|
|
|
|
|
|
"wavernn":
|
|
|
|
|
|
|
|
"paddlespeech.t2s.models.wavernn:WaveRNN",
|
|
|
|
|
|
|
|
"wavernn_inference":
|
|
|
|
|
|
|
|
"paddlespeech.t2s.models.wavernn:WaveRNNInference",
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@ -148,10 +152,16 @@ def evaluate(args):
|
|
|
|
voc_name = args.voc[:args.voc.rindex('_')]
|
|
|
|
voc_name = args.voc[:args.voc.rindex('_')]
|
|
|
|
voc_class = dynamic_import(voc_name, model_alias)
|
|
|
|
voc_class = dynamic_import(voc_name, model_alias)
|
|
|
|
voc_inference_class = dynamic_import(voc_name + '_inference', model_alias)
|
|
|
|
voc_inference_class = dynamic_import(voc_name + '_inference', model_alias)
|
|
|
|
|
|
|
|
if voc_name != 'wavernn':
|
|
|
|
voc = voc_class(**voc_config["generator_params"])
|
|
|
|
voc = voc_class(**voc_config["generator_params"])
|
|
|
|
voc.set_state_dict(paddle.load(args.voc_ckpt)["generator_params"])
|
|
|
|
voc.set_state_dict(paddle.load(args.voc_ckpt)["generator_params"])
|
|
|
|
voc.remove_weight_norm()
|
|
|
|
voc.remove_weight_norm()
|
|
|
|
voc.eval()
|
|
|
|
voc.eval()
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
voc = voc_class(**voc_config["model"])
|
|
|
|
|
|
|
|
voc.set_state_dict(paddle.load(args.voc_ckpt)["main_params"])
|
|
|
|
|
|
|
|
voc.eval()
|
|
|
|
|
|
|
|
|
|
|
|
voc_mu, voc_std = np.load(args.voc_stat)
|
|
|
|
voc_mu, voc_std = np.load(args.voc_stat)
|
|
|
|
voc_mu = paddle.to_tensor(voc_mu)
|
|
|
|
voc_mu = paddle.to_tensor(voc_mu)
|
|
|
|
voc_std = paddle.to_tensor(voc_std)
|
|
|
|
voc_std = paddle.to_tensor(voc_std)
|
|
|
@ -307,7 +317,8 @@ def main():
|
|
|
|
default='pwgan_csmsc',
|
|
|
|
default='pwgan_csmsc',
|
|
|
|
choices=[
|
|
|
|
choices=[
|
|
|
|
'pwgan_csmsc', 'pwgan_ljspeech', 'pwgan_aishell3', 'pwgan_vctk',
|
|
|
|
'pwgan_csmsc', 'pwgan_ljspeech', 'pwgan_aishell3', 'pwgan_vctk',
|
|
|
|
'mb_melgan_csmsc', 'style_melgan_csmsc', 'hifigan_csmsc'
|
|
|
|
'mb_melgan_csmsc', 'style_melgan_csmsc', 'hifigan_csmsc',
|
|
|
|
|
|
|
|
'wavernn_csmsc'
|
|
|
|
],
|
|
|
|
],
|
|
|
|
help='Choose vocoder type of tts task.')
|
|
|
|
help='Choose vocoder type of tts task.')
|
|
|
|
|
|
|
|
|
|
|
|