Oh no. Error running app - but no errors in log

I’ve successfully deployed an app locally but when I run it in the cloud it crashes without error (there is one error in installing dependencies that the protobuf version is not compatible with streamlit/tensorflow but I’ve removed the tf import and the error still happens). From similar topics it seems this might be a memory issue but I’ve stripped the code right back and am still having the same error.

Any pointers or advice on how to reduce/profile memory usage or other ways to figure out the issue would be much appreciated.

Hey @duncanwp, can you share the link to the deployed app?

Sure: https://duncanwp-climatebench-app-streamlit-app-gr3ej2.streamlit.app/

And the code is here: GitHub - duncanwp/climatebench-app

It looks like the app is getting stuck on this error:

tensorflow 2.11.0 requires protobuf<3.20,>=3.9.2, but you have protobuf 3.20.1 which is incompatible.

What happens if you use an older version of protobuf?

I wondered that, but the error is at install time and I removed the tensorflow import which I thought meant it wouldn’t cause an issue at runtime.

As far as I can see this is the version of protobuf needed by streamlit and gets installed by an internal streamlit pip install which overwrites the version I specify in my install script. But I might be misunderstanding the installation workflow…

No, streamlit requires protobuf<4,>=3.12.