atari-representation-learning
atari-representation-learning copied to clipboard
mila-iqia
Code for "Unsupervised State Representation Learning in Atari"
Hello, I was checking the [ram annotations ](https://github.com/mila-iqia/atari-representation-learning/blob/master/atariari/benchmark/ram_annotations.py) for Pong and it seems like only the score and position of the players and ball is known from the ram, my...
(The following is my observation through trial and error. So it may not be 100% correct.) In MsPacman, by default, ram[116] == 0. When MsPacman eats the Power Pellet, and...
I tried to run probing tasks for different Atari environments, using the following command: `python -m scripts.run_probe --method infonce-stdim --env-name {env_name}` I did not change any code, just tried different...
https://github.com/mila-iqia/atari-representation-learning/blob/f5e1080fe2077356b409cd324d6847b03ed35308/atariari/benchmark/ram_annotations.py#L68 In the Berzerk game dictionary there are 8 indices corresponding to robot_x coordinates and 9 indices corresponding to robot_y coordinates. ``` enemy_robots_x=range(65, 73), enemy_robots_y=range(56, 65), ``` Is this a...
I feel the features extracted for seaquest are incorrect/ambiguous. In the attached image, there are 1. 4 enemies with one of them being at a different position than the others...
Awesome work guys. Can't wait to take this for a spin. Recently Mnih's group released a paper on representation learning using Atari. Unsupervised Learning of Object Keypoints for Perception and...
### Episodes **Before:** - list of lists of pytorch tensors (one tensor for each example) **After:** - list of pytorch tensors (one tensor for each episode) ### Labels **Before:** -...
Did you find out how the positions of the sprites are encoded? For example in Asteroids, I want to find the asteroid or the player given its x-y-coordinate in the...
This pull request adds: 1. Room conditional RAM index dictionaries for MZR and Venture 2. Adapts categorization code for these room-conditional dicts 3. Adapts probe code to filter out an...