if args.use_cuda:
model.cuda()
model.decoder.set_r(cp['r'])
# load vocoder model
if args.vocoder_path != "":
VC = load_config(args.vocoder_config_path)
ap_vocoder = AudioProcessor(**VC.audio)
bits = 10
vocoder_model = VocoderModel(rnn_dims=512,
fc_dims=512,
mode=VC.mode,
mulaw=VC.mulaw,
pad=VC.pad,
upsample_factors=VC.upsample_factors,
feat_dims=VC.audio["num_mels"],
compute_dims=128,
res_out_dims=128,
res_blocks=10,
hop_length=ap.hop_length,
sample_rate=ap.sample_rate,
use_aux_net=True,
use_upsample_net=True)
check = torch.load(args.vocoder_path)
vocoder_model.load_state_dict(check['model'])
vocoder_model.eval()
if args.use_cuda:
vocoder_model.cuda()
else:
vocoder_model = None
VC = None
model.cuda()
model.decoder.set_r(cp['r'])
# load vocoder model
if args.vocoder_path != "":
VC = load_config(args.vocoder_config_path)
ap_vocoder = AudioProcessor(**VC.audio)
if VC.bits is not None:
vocoder_model = VocoderModel(
rnn_dims=512,
fc_dims=512,
bits=VC.bits,
pad=VC.pad,
upsample_factors=VC.upsample_factors, # set this depending on dataset
feat_dims=80,
compute_dims=128,
res_out_dims=128,
res_blocks=10,
hop_length=ap.hop_length,
sample_rate=ap.sample_rate
)
else:
vocoder_model = VocoderModel(
rnn_dims=512,
fc_dims=512,
mode=VC.mode,
mulaw=VC.mulaw,
pad=VC.pad,
use_aux_net=VC.use_aux_net,
use_upsample_net=VC.use_upsample_net,