How can we make tensorflow work with this docker image? about uwsgi-nginx-docker HOT 2 CLOSED

Hey! You might be running into timeouts from Nginx or from uWSGI. If the Meinheld image works for you, isn't that better?

In any case, if you are running a TensorFlow model (or any other type of ML with a model that requires a lot of memory) you would probably be better off with a simple Dockerfile built from scratch instead of this (or any other) base image. And using, for example, only Gunicorn with a single worker process.

That way you can manually and carefully increase the number of containers from outside, probably with Kubernetes, Docker Compose, Docker Swarm, or something similar. That way you don't risk having too many processes that consume all the memory and cause problems for the others.

And if you have a single process per container, then all the extra logic from this one is just unnecessary complexity.

from uwsgi-nginx-docker.

Assuming the original issue was solved, it will be automatically closed now. But feel free to add more comments or create new issues.

from uwsgi-nginx-docker.