Updating TensorFlow version on Deepo base image?


Hi there,

I’m running a notebook on Gradient on the Deepo (Python 3.6) base container image and the installed version of TensorFlow is a bit behind (1.8 vs 1.12+).

Is updating the version supported and is there anything extra I should do besides a pip install? I’ve tried updating it on another base image and ran into a version mismatch for some CUDA related libraries.

Also, I think that there is a newer version of the Deepo image in the Docker hub, can I run it directly as a “custom” container image or are there additional requirements for them to be compatible with Gradient?

It’s not a dealbreaker for now, but some minor changes are required to the code I’m running because of API differences.


@supernes Great question. It’s super easy to pull the latest container instead of clicking one of the options on the list of tiles. Just use the Custom Container option like you mentioned and choose whatever container you want from Deepo. Here’s an example pulling the latest Python 3.6 container:

Note: Only the first line is required (Jupyter will start automatically by default).

The path/tag I used in this one is ufoym/deepo:all-jupyter-py36 and here is the list of other tags you can use: https://hub.docker.com/r/ufoym/deepo/tags

Hope that helps!


That’s fantastic, thank you! I just tried it and it works perfectly :+1:t2: