Wdunn001/gpu

[wdunn001]

adding commands from non gpu version

[wdunn001]

resolved this issue now receiving two issues with piper and openwakeword

Piper:

usage: __main__.py [-h] --piper PIPER --voice VOICE [--uri URI] --data-dir
                   DATA_DIR [--download-dir DOWNLOAD_DIR] [--speaker SPEAKER]
                   [--noise-scale NOISE_SCALE] [--length-scale LENGTH_SCALE]
                   [--noise-w NOISE_W] [--auto-punctuation AUTO_PUNCTUATION]
                   [--samples-per-chunk SAMPLES_PER_CHUNK]
                   [--max-piper-procs MAX_PIPER_PROCS] [--update-voices]
                   [--debug]
__main__.py: error: unrecognized arguments: --cuda

OpenWakeWord:

Traceback (most recent call last):
  File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/usr/local/lib/python3.10/dist-packages/wyoming_openwakeword/__main__.py", line 176, in <module>
    asyncio.run(main())
  File "/usr/lib/python3.10/asyncio/runners.py", line 44, in run
    return loop.run_until_complete(main)
  File "/usr/lib/python3.10/asyncio/base_events.py", line 649, in run_until_complete
    return future.result()
  File "/usr/local/lib/python3.10/dist-packages/wyoming_openwakeword/__main__.py", line 87, in main
�
    assert (
AssertionError: Missing model: hey_jarvis
 (looked in: /usr/local/lib/python3.10/dist-packages/wyoming_openwakeword/models)

[wdunn001]

Do I need a PyTorch instance to run piper on gpu?

[edurenye]

About piper, see the issue I created here: rhasspy/rhasspy3#49
I couldn't fix that issue, we would need to build our own python library or just wait for the maintainer of the library to add the support for that, there is no repo where we can make a PR for that sadly

[edurenye]

Sorry for that

[edurenye]

Why did you change it to single line? Seems unrelated and reduces legibility

[wdunn001]

oh I did this because if I don't it won't run the script. its not recognizing the / I think this is a different in my environment refactoring it ensures its environment agnostic. I get a command not recognized error for each new line

[edurenye]

Weird, this is supposed to run inside the container, should not care about your environment.

[edurenye]

Do I need a PyTorch instance to run piper on gpu?

No, according to the documentation, you just need onnxruntime-gpu

[edurenye]

I think preloading the model should fix the opwnwakeword issue. Not sure if we should preload that one, feels the more neutral and easy to pronounce to me, but maybe I'm wrong.

Seems like the maintainer was using ok_nabu, but because 'Nabu Casa' is the company where he works, personally I think the dafault should be something neutral and non comercial word, then they can use it for their products with whatever model they want.

[wdunn001]

I just wanted to update you I cannot get OpenWakeWord going I tried preloading a model but they can't be found. I need to redo the docker file so I can peak into the default folder to see if the model is actually there. What I think might need to happen is we allow the end user to define the model location and provide a ling for download to a list of default models. I haven't had time to set all that up recently.