Protip 2.0, feeling gutsy and wanna belatedly add files after your mass import? be sure to generate a spectrogram for each new wav you add!

What's the problem here?

On the surface, it looks like there's a rogue square bracket in your 'list.txt' file - it's listing 'wavs/1.wav]' as the filename (as opposed to 'wavs/1.wav')

you might have empty lines in your transcript file

Hmmm

please show all of the error msg

did you fix your text file and started the process with the new text file?

it doesn't look like it

I did

thought the .npy it should generate dont seem to exist based on the error msg

and you should really make your dataset larger

19 files isn't enough

also you need to have longer files too

up to about 10 seconds in length

you longest file is 5 sec long

could someone give a quick "speech synthesis NN training for dummies" type guide ?

what software to use / how to load files for training / etc

nvm just checked pinned msgs lol

you can read this chat too

you'll see what the most common problems are and how to fix them

thanks!

I thought I'd pinned that video already but I guess not

best of luck!

Hmm, that's odd. All of a sudden, I'm getting some strange CUDA memory errors:

FP16 Run: False
Dynamic Loss Scaling: True
Distributed Run: False
cuDNN Enabled: True
cuDNN Benchmark: False
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-13-e8a871a0e98d> in <module>()
      5 print('cuDNN Benchmark:', hparams.cudnn_benchmark)
      6 train(output_directory, log_directory, checkpoint_path,
----> 7       warm_start, n_gpus, rank, group_name, hparams, log_directory2)

5 frames
/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py in <lambda>(t)
    489             Module: self
    490         """
--> 491         return self._apply(lambda t: t.cuda(device))
    492 
    493     def xpu(self: T, device: Optional[Union[int, device]] = None) -> T:

RuntimeError: CUDA out of memory. Tried to allocate 2.00 MiB (GPU 0; 15.78 GiB total capacity; 14.42 GiB already allocated; 2.75 MiB free; 14.48 GiB reserved in total by PyTorch)

I've been allocated a Tesla V100, and this happens regardless of the batch size (I've tried 30, 16 and even 1). The thing that strikes me as weird is there is enough memory for allocation (it tries to allocate 2.00 MiB when there is 2.75 MiB free)

just decrease the batch size

I decreased it in stages, all the way down to 1 and it still did it.

can I see your dataset then please