search

dusty-nv / jetson-containers

Machine Learning Containers for NVIDIA Jetson and JetPack-L4T
MIT License
1.89k stars 416 forks source link

fixed not loading audiocraft examples, and added voicecraft dockerfile #498

Closed martincerven closed 2 months ago

martincerven commented 2 months ago

Fixed "Failed to load audio" RuntimeError. - it was probably combination of two things:

  1. old libsndfile that doesn't play mp3
  2. didn't use the e flag
    • normally pip install . installs into/dist-packages/<package>
    • e enables editable mode where it only symlinks it into /dist-packages/ and audiocraft repo uses it with relative paths to get assets and other scripts/programs, that's where it breaks loading of files

Added VoiceCraft container

  1. It is SOTA Voice editing/ TTS
  2. In the repo they use conda, but after checking the issues it is not really needed:
    • the packages they use with conda are for alignment which whisperx does too

I put gradio_app.py in the CMD instruction, since otherwise the jupyter from previous layer/package would run, it runs 0.0.0.0:7860at start.

dusty-nv commented 2 months ago

Thanks @martincerven! Built this and pushed dustynv/voicecraft:r36.2.0, and added voicecraft to jetson-containers main readme under the Audio category, and provided you attribution in the voicecraft readme.

Also rebuilt dustynv/audiocraft:r36.2.0 with your fixes - thanks for digging into that :+1:

If you'd like to be listed on the research group members page, send me your desired photo/avatar, and I'll get you up there!