Closed spMohanty closed 4 years ago
This seems to be causing a lof build failues in the OTC challenge after we upgraded aicrowd-repo2docker
to use nvidia/cuda:10.0-cudnn7-runtime-ubuntu18.04
as the base image.
@spMohanty - ml-agents only supports python 3.6 python_requires=">=3.6,<3.7"
reference https://github.com/Unity-Technologies/ml-agents/blob/master/ml-agents/setup.py
@Sohojoe : Yeah, I saw that. This issue was mostly for extending support forr py37
.
(Though in the short term participants can choose py36 in their runtime specification using aicrowd-repo2docker
, just a bit worried that it might cause some confusion among participants.
Hi @spMohanty
As @Sohojoe says, we are requiring users to have 3.6 as their environment. This has been the case since Round 1, so all participants in Round 2 should know what to do. We make this explicit on the obstacle-tower-env
page. I am happy to add a notice somewhere else in the documentation if you believe it would be helpful. In a future version of Obstacle Tower we also plan to fully support 3.7.
Ran into this too. https://github.com/Unity-Technologies/obstacle-tower-env#requirements mentions 3.5 - 3.7 so that is wrong.
conda create –name=unityml
conda activate unityml
conda install python=3.6
pip install -e .