Issues with running examples text_to_audio_playback.py and gui_text_to_audio_playback.py circular import and webdataset.

[lukaLLM]

from webdataset.filters import default_collation_fn ModuleNotFoundError: No module named 'webdataset'

I solved it just installing the thing but then I got this error.

ImportError: cannot import name 'Pipeline' from partially initialized module 'whisperspeech.pipeline' (most likely due to a circular import) \site-packages\whisperspeech\pipeline.py)

[BBC-Esq]

Hmm...let me look into it.

[BBC-Esq]

I determined that WhisperSpeech needs webdataset for its TTS functionality. Also, I wasn't able to verify entirely...even when I used 0.8 I still encountered the same message. Unless/until this repository is actively maintained again, I'm recommending that people install my fork's version:

You can pip install it with this command:

pip install git+https://github.com/BBC-Esq/WhisperSpeech.git@795f60157136b0052b9a1f576e88803f7783ab1f

[BBC-Esq]

I added webdataset to the instructions in each example file to clarify. This repo does in-fact list it as a dependency but only for "development" purposes. I pip installed version 0.8.9 and am not getting the errors anymore you mentioned.

I'll leave it to the owners of the repository to make webdataset mandatory for all installations, not just "dev" purposes, but in my experience it is in-fact necessary like you say.

Closing for now.

[BBC-Esq]

I actually could modify it. Should be resolved in this:

https://github.com/WhisperSpeech/WhisperSpeech/pull/170