https://github.com/sdpkjc/snapshotrl

Open source code for "Snapshot Reinforcement Learning: Leveraging Prior Trajectories for Efficiency"
https://github.com/sdpkjc/snapshotrl

Last synced: 8 days ago
JSON representation

Open source code for "Snapshot Reinforcement Learning: Leveraging Prior Trajectories for Efficiency"

• Host: GitHub
• URL: https://github.com/sdpkjc/snapshotrl
• Owner: sdpkjc
• Created: 2024-03-11T13:34:59.000Z (8 months ago)
• Default Branch: main
• Last Pushed: 2024-03-14T01:45:18.000Z (8 months ago)
• Last Synced: 2024-10-11T05:50:59.958Z (about 1 month ago)
• Language: Python
• Homepage: https://sdpkjc.github.io/snapshotrl/
• Size: 463 KB
• Stars: 2
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# 🚀 Snapshot Reinforcement Learning: *Leveraging Prior Trajectories for Efficiency*













This repository contains the code for the paper "Snapshot Reinforcement Learning: Leveraging Prior Trajectories for Efficiency" by [Yanxiao Zhao](https://sdpkjc.com), Yangge Qian, Tianyi Wang, Jingyang Shan, Xiaolin Qin.

![snapshotrl](./snapshotrl.svg)

## Running the code

The code is written in Python 3.9 and PyTorch 2.0.1. To install the required dependencies, run

```bash

pip install -r src/requirements.txt

```

To train the agents, run the following commands:

```bash

# for S3RL+TD3

python src/snapshot_td3_mujoco.py --env-id Hopper-v4 --n-clusters 6 --truncate-step 100 --seed 1

# for SnapshotRL+SC+TD3

python src/snapshot_td3_mujoco.py --env-id Hopper-v4 --n-clusters 6 --truncate-step 1000 --seed 1

# for SnapshotRL+STT+TD3

python src/snapshot_td3_mujoco_random.py --env-id Hopper-v4 --truncate-step 100 --seed 1

# for SnapshotRL+TD3

python src/snapshot_td3_mujoco_random.py --env-id Hopper-v4 --truncate-step 1000 --seed 1

# for S3RL+SAC

python src/snapshot_sac_mujoco.py --env-id Hopper-v4 --n-clusters 6 --truncate-step 100 --seed 1

# for S3RL+PPO

python src/snapshot_ppo_fix_mujoco.py --env-id Hopper-v4 --n-clusters 6 --truncate-step 100 --seed 1

```

## Citing this work

```tex

@article{zhao2024snapshot,

  title={Snapshot Reinforcement Learning: Leveraging Prior Trajectories for Efficiency},

  author={Zhao, Yanxiao and Qian, Yangge and Wang, Tianyi and Shan, Jingyang and Qin, Xiaolin},

  journal={arXiv preprint arXiv:2403.00673},

  year={2024}

}

```