CaraDuf

Great! Thanks a lot @blue-fish ! With your guide I should be able to do it!

I am a bit stuck : 1. How to generate the `stop` in `stop[j, :int(dataset.metadata[k][4])-1] = 0` ? 2. I get the following runtime error while calling `loss.backward()` in the...

Thanks @blue-fish . Yet `Only Tensors of floating point and complex dtype can require gradients` appears when passing `requires_grad=true`. If I also pass `dtype=torch.floating32` to get floating point as required...

Ok it works (well actually it does not look like it does not work). It does not bring many more steps / s but it should follow their [recipe](https://pytorch.org/tutorials/recipes/recipes/tuning_guide.html?highlight=device#pre-allocate-memory-in-case-of-variable-input-length). I...

I would also like to see if "pointwise operations" can benefit from the [PyTorch JIT](https://pytorch.org/tutorials/recipes/recipes/tuning_guide.html#fuse-pointwise-operations). Are there such operations ?

This is a great answer, thank you! I could do transfer learning with the provided model and quickly have a French tts. Cool!

@annaklyueva this is what I did (using the restore flag with the model_file provided in the released model) and it worked.

Hi @GerrySant did you try running train-tts.py instead of a recipe as suggested by @erogol (see https://github.com/coqui-ai/TTS/discussions/1719#discussioncomment-3187998)? I believe I was facing the same issue and this trick solved it.

Hi @GerrySant the "name" parameter in the "datasets" section in your config file deals with the formatter name not the voice name. So if you formatted your dataset with an...

@p0p4k now that things are clearer, I will try and write something. According to @WeberJulian in [gitter](https://matrix.to/#/@weberjulian:matrix.org?utm_source=gitter) : > No it was for a time since phonemizer wasn't in a...