mildemelwe / nnsvs-resf0nonattentivetacotron-pretrained

Pretrained model for NNSVS's ResF0NonAttentiveTacotron acoustic model type

• Hi, I’m @Mildemelwe
• I’m interested in text-to-speech software
• I’m currently working on my fork of neutrogic/neuTalk, as well as something secret...
• Reach me via email at [email protected], or through discord at Kei Wendt#6210

nnsvs-resf0nonattentivetacotron-pretrained's Issues

possible bug in duration training

the following error appeared while training duration with my google translate database. cause of error is currently unknown but i sorta suspect it's a bug with the notebook's way of handling duration pretraining. pretraining for all other models worked fine.

Error executing job with overrides: ['+data.in_scaler_path=dump/Google_Translate_ResF0NonAttentiveTacotron/norm/in_duration_scaler.joblib', '+data.out_scaler_path=dump/Google_Translate_ResF0NonAttentiveTacotron/norm/out_duration_scaler.joblib', 'model=duration_vp_mdn', 'train=myconfig', 'data=myconfig', 'data.train_no_dev.in_dir=dump/Google_Translate_ResF0NonAttentiveTacotron/norm/train_no_dev/in_duration/', 'data.train_no_dev.out_dir=dump/Google_Translate_ResF0NonAttentiveTacotron/norm/train_no_dev/out_duration/', 'data.dev.in_dir=dump/Google_Translate_ResF0NonAttentiveTacotron/norm/dev/in_duration/', 'data.dev.out_dir=dump/Google_Translate_ResF0NonAttentiveTacotron/norm/dev/out_duration/', 'data.in_scaler_path=dump/Google_Translate_ResF0NonAttentiveTacotron/norm/in_duration_scaler.joblib', 'data.out_scaler_path=dump/Google_Translate_ResF0NonAttentiveTacotron/norm/out_duration_scaler.joblib', 'train.out_dir=exp/Google_Translate_ResF0NonAttentiveTacotron_intunist_prototyping_notebook/duration', 'train.resume.checkpoint=/content/exp/ResF0NonAttentiveTacotron_pretrained_intunist_prototyping_notebook/duration/best_loss.pth']
Traceback (most recent call last):
  File "/usr/local/lib/python3.7/dist-packages/nnsvs/bin/train.py", line 257, in my_app
    ) = setup(config, device, collate_fn)
  File "/usr/local/lib/python3.7/dist-packages/nnsvs/train_util.py", line 513, in setup
    model.load_state_dict(checkpoint["state_dict"])
  File "/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py", line 1605, in load_state_dict
    self.__class__.__name__, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for VariancePredictor:
	Missing key(s) in state_dict: "conv.0.0.weight", "conv.0.0.bias", "conv.0.2.weight", "conv.0.2.bias", "conv.1.0.weight", "conv.1.0.bias", "conv.1.2.weight", "conv.1.2.bias", "conv.2.0.weight", "conv.2.0.bias", "conv.2.2.weight", "conv.2.2.bias", "conv.3.0.weight", "conv.3.0.bias", "conv.3.2.weight", "conv.3.2.bias", "mdn_layer.log_pi.weight", "mdn_layer.log_pi.bias", "mdn_layer.log_sigma.weight", "mdn_layer.log_sigma.bias", "mdn_layer.mu.weight", "mdn_layer.mu.bias". 
	Unexpected key(s) in state_dict: "model.0.weight", "model.0.bias", "model.3.weight", "model.3.bias", "model.6.weight", "model.6.bias", "model.9.log_pi.weight", "model.9.log_pi.bias", "model.9.log_sigma.weight", "model.9.log_sigma.bias", "model.9.mu.weight", "model.9.mu.bias". 

Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.