I'm deploying a webapp on Google Cloud Plateform to test some trained models with Fast.ai. One of them is for sound recognition, and I need to use librosa to create a spectogram of the sound provided by the user. But librosa needs ffmpeg to work properly.
I added ffmpeg to my Dockerfile. It works fine when I deploy the application, but after few requests I get this error (as if ffmpeg is not installed) :
File "/usr/local/lib/python3.6/site-packages/audioread/init.py", line 116, in audio_open: raise NoBackendError() at load (/usr/local/lib/python3.6/site-packages/librosa/core/audio.py:119)
Sometimes, it works again. It looks like it depends on which instance the application is running on.
Here is my Dockefile :
FROM python:3.6-slim-stretch
RUN apt update
RUN apt install -y python3-dev gcc
RUN apt install -y ffmpeg
ADD requirements.txt requirements.txt
RUN pip install -r requirements.txt
COPY app app/
RUN python app/server.py
EXPOSE 8080
CMD ["python", "app/server.py", "serve"]```