Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/ethanabrooks/decision-transformer
https://github.com/ethanabrooks/decision-transformer
Last synced: 1 day ago
JSON representation
- Host: GitHub
- URL: https://github.com/ethanabrooks/decision-transformer
- Owner: ethanabrooks
- License: mit
- Created: 2022-06-25T18:08:42.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2022-06-30T16:33:48.000Z (over 2 years ago)
- Last Synced: 2023-08-18T13:42:29.123Z (about 1 year ago)
- Language: Python
- Size: 373 KB
- Stars: 1
- Watchers: 4
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Atari
We build our Atari implementation on top of [minGPT](https://github.com/karpathy/minGPT) and benchmark our results on the [DQN-replay](https://github.com/google-research/batch_rl) dataset.
## Installation
Dependencies can be installed with the following command:
```
conda env create -f conda_env.yml
```
## Downloading datasets
Create a directory for the dataset and load the dataset using [gsutil](https://cloud.google.com/storage/docs/gsutil_install#install). Replace `[DIRECTORY_NAME]` and `[GAME_NAME]` accordingly (e.g., `./dqn_replay` for `[DIRECTORY_NAME]` and `Breakout` for `[GAME_NAME]`)
```
mkdir [DIRECTORY_NAME]
gsutil -m cp -R gs://atari-replay-datasets/dqn/[GAME_NAME] [DIRECTORY_NAME]
```
## Example usage
Scripts to reproduce our Decision Transformer results can be found in `run.sh`.
```
python run_dt_atari.py --seed 123 --block_size 90 --epochs 5 --model_type 'reward_conditioned' --num_steps 500000 --num_buffers 50 --game 'Breakout' --batch_size 128 --data_dir_prefix [DIRECTORY_NAME]
```