search

Simsso / NIPS-2018-Adversarial-Vision-Challenge

Code, documents, and deployment configuration files, related to our participation in the 2018 NIPS Adversarial Vision Challenge "Robust Model Track"
MIT License
10 stars 1 forks source link

Python 3.6 Compatibility #66

Closed Simsso closed 5 years ago

Simsso commented 5 years ago

Right now our pipeline supports only Python 3.5 because it is baked into Ubuntu 16 and the tensorflow:latest-gpu-py3 Docker image. However, for attribute type annotations and simply in order to be at least almost up to date (Python 3.7 is not supported by TensorFlow), it would be nice to support Python 3.6.

@FlorianPfisterer suggested using a virtual environment, which seems feasible. What's the right place to add the command line commands for setting up the venv and installing things, @doktorgibson? Is it the Dockerfile or some bash script?

I'd say we support Python 3.6 unless it's more than 30 min of work. Maybe we can do this together real quick. It involves undoing this commit https://github.com/Simsso/NIPS-2018-Adversarial-Vision-Challenge/commit/e91d242ccb0d553558a9f69cdeabe7a4db78c35b.

ghost commented 5 years ago

I still don't see any real benefits to our work when using Python3.6. Upon official release you can easily switch to 3.6 annotations, since at that moment you wouldn't use the pipeline anymore.

If you propose a fast and clean solution I can add it to the pipeline, otherwise I would like to invest my time into other WI.

ghost commented 5 years ago

Currently I don't have any clue about how things work in Python. So completely amateur in this point :/

The right place is at the start.sh, since you don't want to bake temporary floating venv into the image.

Simsso commented 5 years ago

Decision made in #56 not to pursue this.