/wasserstein_actor_critic

Code for the paper "Wasserstein Actor-Critic: Directed Exploration via Optimism for Continuous-Actions Control"

Primary LanguagePython

Wasserstein Actor Critic

This repository contains the code accompanying the AAAI 2023 paper 'Wasserstein Actor-Critic: Directed Exploration via Optimism for Continuous-Actions Control'. This is a fork of https://github.com/microsoft/oac-explore

Running Experiments

For software dependencies, please have a look inside the requirements.txt

To run Soft Actor Critic on Humanoid with seed 0 as a baseline to compare against Optimistic Actor Critic, run

python main.py --seed=0 --domain=humanoid

To run Wasserstein Actor Critic on Point 4 with random seed and the hyperparameters from the paper launch:

python main.py --seed=0 --domain=point --terminal --clip_state --difficulty=double_L --alg=gs-oac --stable_critic --delta=0.95 --max_path_length=300 --std_lr=1e-3 --policy_lr=1e-3 --qf_lr=1e-3 --r_min=-0.5 --r_max=0.5 --prv_std_qty=0.6 --prv_std_weight=0.6 --epochs=300 --no_gpu --suffix=DL3_GSWAC

The same file can be used to reproduce all the results, with all the environments and algorithms, by modyfying the parameter values.