[BUG] Segmentation fault on Intel Macbook

[ybwai]

Bug Description Running RVC single infer (convert_audio) on an Intel based Macbook pro with PYTORCH_ENABLE_MPS_FALLBACK=1 does not seem to be working. Get a segmentation fault after a minute or two.

Steps to Reproduce Outline the steps to replicate the issue:

1. I created a simple script:

   
   import os
   from rvc.infer.infer import VoiceConverter

converter = VoiceConverter() converter.convert_audio( audio_input_path=str(input_path), audio_output_path=str(output_path), model_path=f"{voice_model_root}/model.pth", index_path=f"{voice_model_root}/model.index", pitch=0, filter_radius=3, index_rate=0.3, volume_envelope=1, protect=0.33, hop_length=128, f0_method="rmvpe", split_audio=False, f0_autotune=False, clean_audio=False, clean_strength=0.7, export_format="WAV", upscale_audio=False, f0_file=None, embedder_model="contentvec", embedder_model_custom=None, formant_shifting=False, formant_qfrency=1.0, formant_timbre=1.0, post_process=False, reverb=False, pitch_shift=False, limiter=False, gain=False, distortion=False, chorus=False, bitcrush=False, clipping=False, compressor=False, delay=False, sliders=additional_params, )

2. I run it with `PYTORCH_ENABLE_MPS_FALLBACK=1 ./.venv/bin/python -m sandbox.rvc`
3. I get this output with a segmentation fault after a couple of minutes:

/.venv/lib/python3.10/site-packages/transformers/tokenization_utils_base.py:1601: FutureWarning: clean_up_tokenization_spaces was not set. It will be set to True by default. This behavior will be depracted in transformers v4.45, and will be then set to False by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884 warnings.warn( Converting audio '/sandbox/sample.mp3'... [1] 9453 segmentation fault PYTORCH_ENABLE_MPS_FALLBACK=1 ./.venv/bin/python -m sandbox.rvc

**Desktop Details:**
- Operating System: MacOS
- Browser: None, on the Terminal (CLI)
- Python v3.10

[blaisewf]

if you buy me a mac maybe i can help you 😂 unfortunately i can't test those environments

[ybwai]

no worries, any way of making it bit more verbose about what is happening and where it might be dying? will try the same script on a unix instance later. does your library usually work without cuda gpus? thanks @blaisewf

[blaisewf]

on inference yes, training requires usually a gpu