kaituoxu/tasnet

A PyTorch implementation of Time-domain Audio Separation Network (TasNet) with Permutation Invariant Training (PIT) for speech separation.

tasnet's Issues

当菜鸟（我）使用高版本torch时，遇到的需要注意的地方 Problems I have met when I using high version torch

1. TypeError: lstm() received an invalid combination of arguments:
   https://blog.csdn.net/yhoon0101/article/details/131649166
   修改代码为：(line 37 in train.py)
   parser.add_argument('--bidirectional', default=True, type=bool, help='Whether use bidirectional LSTM')

2. RuntimeError: ‘lengths‘ argument should be a 1D CPU int64 tensor, but got 1D cuda:0 Long tensor:
   https://blog.csdn.net/weixin_45084253/article/details/122269699
   修改代码为：(line 179 in solver.py)
   estimate_source = self.model(padded_mixture, mixture_lengths.to("cpu"))

我升级为高版本torch是因为遇到runtimeerror: cudnn error: cudnn_status_mapping_error错误，我的原因是因为torch版本和cuda版本不太兼容

didn't match because some of the arguments have invalid types: (Tensor, Tensor, !tuple!, !list!, bool, int, int, bool, !int!)

Hello, thanks for sharing your code. I'm very interested in the project. But while running the project, I encountered a problem:

Training...
Traceback (most recent call last):
  File "~/TasNet/egs/wsj0/../../src/train.py", line 125, in <module>
    main(args)
  File "~/TasNet/egs/wsj0/../../src/train.py", line 119, in main
    solver.train()
  File "~/TasNet/src/solver.py", line 75, in train
    tr_avg_loss = self._run_one_epoch(epoch)
  File "~/TasNet/src/solver.py", line 174, in _run_one_epoch
    estimate_source = self.model(padded_mixture, mixture_lengths)
  File "~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)
  File "~/TasNetsrc/tasnet.py", line 36, in forward
    est_mask = self.separator(mixture_w, mixture_lengths)
  File "~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)
  File "~/TasNet/src/tasnet.py", line 155, in forward
    packed_output, hidden = self.rnn(packed_input)
  File "~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)
  File "~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py", line 182, in forward
    self.num_layers, self.dropout, self.training, self.bidirectional)
TypeError: lstm() received an invalid combination of arguments - got (Tensor, Tensor, tuple, list, bool, int, int, bool, int), but expected one of:
 * (Tensor data, Tensor batch_sizes, tuple of Tensors hx, tuple of Tensors params, bool has_biases, int num_layers, float dropout, bool train, bool bidirectional)
      didn't match because some of the arguments have invalid types: (Tensor, Tensor, !tuple!, !list!, bool, int, int, bool, !int!)
 * (Tensor input, tuple of Tensors hx, tuple of Tensors params, bool has_biases, int num_layers, float dropout, bool train, bool bidirectional, bool batch_first)
      didn't match because some of the arguments have invalid types: (Tensor, !Tensor!, !tuple!, !list!, !bool!, int, !int!, bool, !int!)

请问您有遇到过这个问题吗？谢谢

您好！想问一下这个是什么错误，不太会

Traceback (most recent call last):
File "G:/speech_code/TasNet/src/train.py", line 56, in
main(ts)
File "G:/speech_code/TasNet/src/train.py", line 52, in main
solver.train()
File "G:\speech_code\TasNet\src\solver.py", line 75, in train
tr_avg_loss = self._run_one_epoch(epoch)
File "G:\speech_code\TasNet\src\solver.py", line 176, in _run_one_epoch
estimate_source = self.model(padded_mixture, mixture_lengths)
File "D:\Users\15229\anaconda3\envs\speech\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "G:\speech_code\TasNet\src\tasnet.py", line 37, in forward
est_mask = self.separator(mixture_w, mixture_lengths)
File "D:\Users\15229\anaconda3\envs\speech\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "G:\speech_code\TasNet\src\tasnet.py", line 162, in forward
packed_output, hidden = self.rnn(packed_input)
File "D:\Users\15229\anaconda3\envs\speech\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "D:\Users\15229\anaconda3\envs\speech\lib\site-packages\torch\nn\modules\rnn.py", line 664, in forward
result = _VF.lstm(input, batch_sizes, hx, self._flat_weights, self.bias,
TypeError: lstm() received an invalid combination of arguments - got (Tensor, Tensor, tuple, list, bool, int, float, bool, int), but expected one of:

• (Tensor data, Tensor batch_sizes, tuple of Tensors hx, tuple of Tensors params, bool has_biases, int num_layers, float dropout, bool train, bool bidirectional)
  didn't match because some of the arguments have invalid types: (Tensor, Tensor, !tuple!, !list!, bool, int, float, bool, !int!)
• (Tensor input, tuple of Tensors hx, tuple of Tensors params, bool has_biases, int num_layers, float dropout, bool train, bool bidirectional, bool batch_first)
  didn't match because some of the arguments have invalid types: (Tensor, !Tensor!, !tuple!, !list!, !bool!, int, !float!, bool, !int!)

Output of speaker order wrong?

I found the label is [spk1,spk2],but output is [spk2,spk1].
can you help me to understand this question?

mixture_w, norm_coef = encoder(mixture)

mixture_w, norm_coef = encoder(mixture)这条语句报错了，encoder没啥用吧

loss becomes negative

What loss function is used and why the loss becomes negative?
Thanks for your nice code.