chenming6615
chenming6615
Maybe it is constrained by the `max_seq_len: 1000` parameter. https://github.com/ming024/FastSpeech2/blob/d4e79eb52e8b01d24703b2dfc0385544092958f3/config/LJSpeech/model.yaml?_pjax=%23js-repo-pjax-container%2C%20div%5Bitemtype%3D%22http%3A%2F%2Fschema.org%2FSoftwareSourceCode%22%5D%20main%2C%20%5Bdata-pjax-container%5D#L33 Can you try to increase it ?
Why do we need `batch_size=batch_size * group_size`? What is the point of the `group_size` variable?
> @chenming6615 Because in the collate_fn function, we need to sort all the utterances in a batch, which will cost time. If we set group_size > 1, we can sort...
Maybe your lexicon generated by MFA contains some phones that are not included in the phone list in the `text/pinyin.py` file. ``` python finals = [ "a1", "a2", "a3", "a4",...
Hi, I investigated this issue by debugging the source code of CPython with GDB. Here is the function in ceval.c that CPython decides a function call should whether be traced...
> > Maybe your lexicon generated by MFA contains some phones that are not included in the phone list in the `text/pinyin.py` file. > > ```python > > finals =...