search

dusty-nv / jetson-containers

Machine Learning Containers for NVIDIA Jetson and JetPack-L4T
MIT License
2.18k stars 446 forks source link

XTTS build #511

Closed budzianowski closed 4 months ago

budzianowski commented 4 months ago

Thanks for the great work! Three questions:

  1. For the version build for 35.2.1 do I have to rebuild python3.9 (this is coqui requirement)? For JP5 what's the best way to go about it?
  2. https://github.com/dusty-nv/jetson-containers/blob/aef3d09267c0b8564c6f3c18dc59873ce0a179d4/packages/audio/xtts/test_voices.py#L12C1-L12C57 - the path to the models have been changed (https://huggingface.co/coqui/XTTS-v2/tree/main)
  3. If I understand the quantization happens during the setup?
dusty-nv commented 4 months ago

@budzianowski I'm not sure that I ever got it working on JetPack 5 and only enable it in my stuff on JetPack 6, and even so found XTTS to be too slow in practice (even on AGX), so I changed to using Piper TTS.

budzianowski commented 4 months ago

Thanks for the heads up - I guess the <200ms latency claims were exaggerated. I will have a look at Piper.

Tetsuo-tek commented 4 months ago

Thanks for jetson-containers, it is very useful!

I would-like to run also xtts container on my jetson-orin-AGX.

$ jetson-containers run $(autotag xtts) Namespace(packages=['xtts'], prefer=['local', 'registry', 'build'], disable=[''], user='dustynv', output='/tmp/autotag', quiet=False, verbose=False) -- L4T_VERSION=36.2.0 JETPACK_VERSION=6.0 CUDA_VERSION=122.0 -- Finding compatible container image for ['xtts']

Couldn't find a compatible container for xtts, would you like to build it? [y/N] y -- Package onnxruntime:1.17 has missing dependencies, disabling... ("couldn't find package: cuda") -- Package onnxruntime:1.17-builder has missing dependencies, disabling... ("couldn't find package: cuda") -- Package tensorflow2 has missing dependencies, disabling... ("couldn't find package: cuda")

[..]

and more and more dependencies not found.

I need to use xtts_v2, and I cannot substitute it with piper. I'm new to jetson and docker. Can you help me to build this container or how to go on the right way?