i trained a michael afton model

Got this error trying to train. Anyone know or is it something in the nb I screwed up?

Python
Exception raised while training: integer division or modulo by zero
Traceback (most recent call last):
  File "/usr/lib/python3.7/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/usr/lib/python3.7/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/usr/local/lib/python3.7/dist-packages/uberduck_ml_dev/exec/train_tacotron2.py", line 49, in <module>
    run(None, None, hparams)
  File "/usr/local/lib/python3.7/dist-packages/uberduck_ml_dev/exec/train_tacotron2.py", line 30, in run
    raise e
  File "/usr/local/lib/python3.7/dist-packages/uberduck_ml_dev/exec/train_tacotron2.py", line 26, in run
    trainer.train()
  File "/usr/local/lib/python3.7/dist-packages/uberduck_ml_dev/trainer/tacotron2.py", line 554, in train
    if epoch % self.epochs_per_checkpoint == 0:
ZeroDivisionError: integer division or modulo by zero

oh wait

-_-

Is there any package/notebook that could automatically separate high quality training audio from low quality training audio? By "quality", I'm referring to how much buzz, noise, artifacts, and other desirable audio content there is in the audio file. Thanks!

why when i train my model more it becomes worse?

You're probably overfitting it in that case.

that is to say, training it for so long that it starts to hyperfocus on exactly the lines you gave it as training data, at the cost of everything else

like a man stuck in a box for 2 weeks starts seeing horses on the coat hook on the wall

Can someone explain this error to me?

FP16 Run: False Dynamic Loss Scaling: True Distributed Run: False cuDNN Enabled: True cuDNN Benchmark: False --------------------------------------------------------------------------- FileNotFoundError Traceback (most recent call last) in 7 print('cuDNN Benchmark:', hparams.cudnn_benchmark) 8 train(output_directory, log_directory, checkpoint_path, ----> 9 warm_start, n_gpus, rank, group_name, hparams, log_directory2) 3 frames /content/tacotron2/utils.py in load_filepaths_and_text(filename, split) 17 18 def load_filepaths_and_text(filename, split="|"): ---> 19 with open(filename, encoding='utf-8') as f: 20 filepaths_and_text = [line.strip().split(split) for line in f] 21 return filepaths_and_text FileNotFoundError: [Errno 2] No such file or directory: 'filelists/clipper_train_filelist.txt'

this error wont let me train the voice

make sure the file that includes the path text pairs are exists this file is missing filelists/clipper_train_filelist.txt

How do I get it? I'm not really good at this whole computer stuff so can you simplify it for me?

could u share the repo link, or the code you're using to help u?

I can't because I'm at work right now. Second. What code would you want me to send? The one with the training cell or the wav conversion cell?

are you using nvidia's repo? this one https://github.com/NVIDIA/tacotron2

?

What is that? How do I check?

When I get a moment when I get home I will

@Mahmoud Salhab I'm back, what do I do?

can anyone help?

i started training and it finished within 11 seconds

i dont know what i did wrong

training on pipeline btw

AttributeError: module 'tensorflow' has no attribute 'contrib' what do I do to fix this?

@hecko I'm currently at work right now and I don't get done til 7. If you're still on and not asleep ill try and get a hold of you again about my problem

what time is it for you right now

1:16pm

ah i'll be awake but not at my pc by then