Closed parthraghav closed 10 months ago
Thank you for your issue. We found the following entry in the FAQ which you may find helpful:
Feel free to close this issue if you found an answer in the FAQ.
If your issue is a feature request, please read this first and update your request accordingly, if needed.
If your issue is a bug report, please provide a minimum reproducible example as a link to a self-contained Google Colab notebook containing everthing needed to reproduce the bug:
Providing an MRE will increase your chance of getting an answer from the community (either maintainers or other power users).
Companies relying on pyannote.audio
in production may contact me via email regarding:
This is an automated reply, generated by FAQtory
I am afraid I won't be able to help you as I have very litte experience with docker
(for now...)
📣 To companies that have successfully deployed pyannote
pipelines in production, now would be the right time to chime in and help @parthraghav.
I can't speak specifically as to why you're seeing a problem, but I can share the configuration that's been working fine in Docker for me:
Environment | Name | Version |
---|---|---|
Docker image | python:3.10-slim | |
pyannote.audio | 3.0.1 | |
torch | 2.0.1 | |
torchaudio | 2.0.2 |
Using Pytorch for CUDA 11.8
I had the same issue.
I was able to run torch==2.1.0
and torchaudio==2.1.0
with pyannote.audio==3.0.0
on my local Mac M2. I also was originally building from python:3.11
.
I down-graded to torch==2.0.1
and torchaudio==2.0.2
with the same PyAnnote Audio version and was able to call Pipeline.from_pretrained
within Docker without the original seg-fault issue. I also build with python:3.10-slim
, so more compatibility testing can be done.
Thank you @hbredin @WhitWaldo @davidthuman
After rebuilding the project requirements from scratch in the docker environment, I learned that pyannote.audio==3.0.1
appears to work in a linux cloud instance and pyannote.audio==3.0.0
works on M1. I introduced separate requirement files for both x86 and arm64 (Apple Silicon) environments.
(base) parthraghav@Parths-Mac-mini worker % cat requirements-x86.txt| grep pyannote
pyannote.audio==3.0.1
pyannote.core==5.0.0
pyannote.database==5.0.1
pyannote.metrics==3.2.1
pyannote.pipeline==3.0.1
(base) parthraghav@Parths-Mac-mini worker % cat requirements-arm64.txt| grep pyannote
pyannote.audio==3.0.0
pyannote.core==5.0.0
pyannote.database==5.0.1
pyannote.metrics==3.2.1
pyannote.pipeline==3.0.1
Hello,
I encountered a problem when trying to load the pretrained
pyannote/speaker-diarization-3.0
model from the pyannote.audio pipeline in therockylinux9/python3
environment. Specifically, I get a Segmentation Fault (SIGSEGV). Interestingly, I can load the model successfully on my Mac M1.Environment:
Error Traceback:
Code to Reproduce:
To reproduce the bug:
cd
into the cloned repo.docker-compose up --no-start --build
.docker exec -it [container_id_or_name] /bin/bash
cmd.python3 audio.py
GDB Backtrace