Unity Obstacle Tower [repo link]
- torch
- torchvision
- gym
- matplotlib
- mlagents (v0.10.1)
- pillow
- opencv-python
conda env create -f environment.yml -n assignment
- The MyAgent will be evaluated
- Initialize model within the
__init__
method - Return action from
act(observation)
method
-
When initializing ObstacleTowerEvaluation env, set docker_training=True
cd to_root_folder docker build -t assignment . docker run -it --rm assignment