google-deepmind / google-deepmind/scalable_agent

Reproducing result for Atari game

Open
#56 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
1k
Forks
160
PR merge metrics
No merged PRs in 30d

Description

Hello,

I am currently trying to apply the IMPALA algorithm to Pong-v0 environment.

I first test my own V-Trace code to Cartpole-v0 and obtain maximum score(https://github.com/kimbring2/minecraft_ai/blob/master/CartPole-v0_IMPALA.ipynb).

The problem is that when I use the same code to Pong-v0(https://github.com/kimbring2/minecraft_ai/blob/master/Pong-v0_IMPALA.ipynb), I can not obtain the score of mentioned at IMPALA paper yet.

It usually take the 1 hour to reach maximum score when I use normal A2C algorithm.

I assume I should change the parameter. Can you give me some hint about that?

Thank you

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.