Install problem on M1

[PierreVannier]

Hi,

It seems this error is related to #1306. I'm running an M1 pro with the following pip list => pyannote.audio 2.1.1 pyannote.core 4.5 pyannote.database 4.1.3 pyannote.metrics 3.2.1 pyannote.pipeline 2.3 torch 2.0.0

python --version
Python 3.10.10

pip install pyannote-audio
$python
Python 3.10.10 | packaged by conda-forge | (main, Mar 24 2023, 20:12:31) [Clang 14.0.6 ] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from pyannote.audio import Pipeline
/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torchaudio/_internal/module_utils.py:99: UserWarning: Failed to import soundfile. 'soundfile' backend is not available.
  warnings.warn("Failed to import soundfile. 'soundfile' backend is not available.")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/pyannote/audio/__init__.py", line 29, in <module>
    from .core.inference import Inference
  File "/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/pyannote/audio/core/inference.py", line 34, in <module>
    from pyannote.audio.core.io import AudioFile
  File "/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/pyannote/audio/core/io.py", line 38, in <module>
    import torchaudio
  File "/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torchaudio/__init__.py", line 1, in <module>
    from torchaudio import (  # noqa: F401
  File "/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torchaudio/_extension.py", line 135, in <module>
    _init_extension()
  File "/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torchaudio/_extension.py", line 105, in _init_extension
    _load_lib("libtorchaudio")
  File "/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torchaudio/_extension.py", line 52, in _load_lib
    torch.ops.load_library(path)
  File "/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torch/_ops.py", line 643, in load_library
    ctypes.CDLL(path)
  File "/Users/pierrevannier/miniforge3/lib/python3.10/ctypes/__init__.py", line 374, in __init__
    self._handle = _dlopen(self._name, mode)
OSError: dlopen(/Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torchaudio/lib/libtorchaudio.so, 0x0006): Symbol not found: __ZN2at4_ops9_pad_enum4callERKNS_6TensorEN3c108ArrayRefIxEExNS5_8optionalIdEE
  Referenced from: <BE6EA463-573E-3D7C-9087-E19500344FCD> /Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torchaudio/lib/libtorchaudio.so
  Expected in:     <78DA4C67-8BC7-3995-8324-D2FD95E57BA4> /Users/pierrevannier/miniforge3/lib/python3.10/site-packages/torch/lib/libtorch_cpu.dylib

[github-actions[bot]]

We found the following entry in the FAQ which you may find helpful:

• Does pyannote support streaming speaker diarization?

Feel free to close this issue if you found an answer in the FAQ. Otherwise, please give us a little time to review.

This is an automated reply, generated by FAQtory

[chainyo]

Hi, @PierreVannier I'm facing the same problem using Poetry and Python 3.10.* on an M1 Mac.

Until it's fully compatible, I would stick to Python 3.9 and Torch 1.13. Because pyannote.audio has a strict version constraint: torchaudio >=0.10,<1.0

[PierreVannier]

Hi, @PierreVannier I'm facing the same problem using Poetry and Python 3.10.* on an M1 Mac.

Until it's fully compatible, I would stick to Python 3.9 and Torch 1.13. Because pyannote.audio has a strict version constraint: torchaudio >=0.10,<1.0

Thanks @chainyo ! I got many troubles on M1 install (even with strict torch version check), it’s causing errors when importing pipelines (after a correct install). Also hmmlearn isn’t installing properly… I guesse M1 isn’t the way to go for this module 😔

[mrmachine]

@PierreVannier On M1, I had problems with torch>=2 and hmmlearn. I was able to get my project running by changing the requirements to install torch<2 (thanks to @chainyo above) and hmmlearn main branch from git:

-e git+https://github.com/hmmlearn/hmmlearn.git#egg=hmmlearn

See: https://github.com/hmmlearn/hmmlearn/issues/514#issuecomment-1494179213

[monaqa]

Package hmmlearn 0.3.0 has been released, which will solve the issue about installing on M1 Mac. However, pyannote-audio 2.1.1 seems to require hmmlearn < 0.3, so installing pyannote-audio still does not work. Why not loosen the version constraints?

https://github.com/pyannote/pyannote-audio/blob/master/requirements.txt#L4

[cwazuidema]

I can confirm that this fix works on M1 Pro. My issue was with hmmlearn as well.

[martinjurkovic]

Since the pull request is still not closed I was able to install pyannote.audio with pip install -c constraints.txt pyannote.audio, where I put hmmlearn>=0.3.0 in the constraints.txt.

[tomklaverpon]

Thanks Farissoliman, this fixed the issue (also for M2 chips). For those using poetry:

pyannote-audio = {git = "https://github.com/Farissoliman/pyannote-audio.git", rev = "patch-1"}

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.