search

Jiwonjeon9603 / MASER

This repository is an implementation of "MASER: Multi-Agent Reinforcement Learning with Subgoals Generated from Experience Replay Buffer" accepted to ICML 2022.
20 stars 7 forks source link

battle won mean always be 0 #4

Closed issuey closed 1 year ago

issuey commented 1 year ago

I use the command like the instrument shows: image

but the battle won mean always be 0 WS0)8V{B)X~D@(RME3K`AZF

why this happen?

Jiwonjeon9603 commented 1 year ago

Did you set the StarCraft environment reward setting for sparse_reward=True?

As you can see in my paper, I used a semi-sparse reward different from the SC2 environment's sparse_reward.