stitching-is-combinatorial-generalisation

KO | EN

gitlite — search

#react #machine-learning #rust #deep-learning #linux #macos #windows #awesome #cli #kubernetes #awesome-list #pytorch #security #openai

stitching-is-combinatorial-generalisation

(★ 25)

[ICLR 2024] Closing the Gap between TD Learning and Supervised Learning - A Generalisation Point of View.

stitching-is-combinatorial-generalisation 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# [Closing the Gap between TD Learning and Supervised Learning -- A Generalisation Point of View](https://arxiv.org/abs/2401.11237)
[Raj Ghugare](https://rajghugare19.github.io/), $\quad$ [Matthieu Geist](https://scholar.google.com/citations?user=ectPLEUAAAAJ), $\quad$ [Glen Berseth](https://neo-x.github.io/)\*, $\quad$ [Benjamin Eysenbach](https://ben-eysenbach.github.io/)\*

\* Equal advising.

## Installation

Create virtual environment named `env_stuff` using command: 
```sh
python3 -m venv env_stuff
```

Install all the packages used to run the code using the `requirements.txt` file: 
```sh
pip install -r requirements.txt
```

## Training

To train an RvS (decision-mlp) agent on pointmaze-umaze using temporal data augmentation, with $\epsilon=0.5$ and $K=40$: 
```sh
python train_dmlp.py dataset_name=pointmaze-umaze-v0 augment_data=True nclusters=40
```

To train a DT (decision-transformer) agent on pointmaze-umaze using temporal data augmentation, with $\epsilon=0.5$ and $K=40$: 
```sh
python train_dt.py dataset_name=pointmaze-umaze-v0 augment_data=True nclusters=40
```

## Datasets

To download the pretrained datasets, visit [this google drive link](https://drive.google.com/drive/folders/1j8Ok2UMYSqfIQReuE6csf1nMoI1s25K-?usp=sharing).

To collect the pointmaze-large dataset with $1e^6$ transitions and seed 1: 
```sh
python collect_pointmaze_data.py pointmaze-large-v0 1 1000000
```

To collect the antmaze-large dataset with $1e^6$ transitions and seed 1: 
```sh
python collect_antmaze_data.py antmaze-umaze-v0 1 1000000
```

## Acknowledgment
Our codebase has been build using/on top of the following codes. We thank the respective authors for their awesome contributions.
- [NanoGPT](https://github.com/karpathy/nanoGPT) 
- [min-decision-transformer](https://github.com/nikhilbarhate99/min-decision-transformer)

## Correspondence

If you have any questions or suggestions, please reach out to me via raj.ghugare@mila.quebec.

stitching-is-combinatorial-generalisation

stitching-is-combinatorial-generalisation 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

stitching-is-combinatorial-generalisation

stitching-is-combinatorial-generalisation 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README