# coreylynch/async-rl

Tensorflow + Keras + OpenAI Gym implementation of 1-step Q Learning from  "Asynchronous Methods for Deep Reinforcement Learning"

Repository: https://github.com/coreylynch/async-rl
Canonical: https://ross.abutalabs.com/products/async-rl
Language: Python
License: MIT
License Family: permissive
Last push: 2018-03-18T04:55:56+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3794, "days_push": 3090, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1003, forks 168 (observed 2026-09-03T02:15:10.215496+00:00)

## What it is
A TensorFlow + Keras implementation of asynchronous 1-step Q-learning from the DeepMind 'Asynchronous Methods for Deep Reinforcement Learning' (A3C) paper, using OpenAI Gym's Atari environments. It runs multiple actor-learner threads to stabilize training without experience replay, and includes a work-in-progress async advantage actor-critic implementation.

## Use cases
- train a deep Q-network on Atari games with async actor-learner threads
- reproduce the asynchronous 1-step Q-learning results from the A3C paper
- learn how to implement deep reinforcement learning with TensorFlow and Keras
- evaluate trained RL agents against OpenAI Gym Atari environments
- run reinforcement learning experiments on low-memory hardware without experience replay
- visualize RL training curves like episode rewards and max Q values in TensorBoard

## When to choose
- you want a compact, readable reference implementation of async Q-learning on Atari
- you're studying the A3C paper and want code to follow along with
- you need RL training that fits in a few GB of RAM without experience replay

## When to avoid
- you need a maintained library - the last release was 2018 and it targets old TensorFlow/Keras versions
- you want production-grade or modern RL tooling like Stable-Baselines3 or RLlib
- you need A3C/actor-critic fully implemented - that part is marked work-in-progress

## Facets
- artifact type: library
- maturity: abandoned
- function: machine-learning, reinforcement-learning, deep-learning, benchmarking
- domain: reinforcement-learning, machine-learning, deep-learning, gaming-tools
- platform: python, cross-platform
- tags: reinforcement-learning, q-learning, a3c, atari, openai-gym, actor-learner, tensorflow, keras, research-code, linux, macos

## Member repositories
- coreylynch/async-rl (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-09-03T02:15:10.215496+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:13:50.151686+00:00, confidence not recorded.
  - readme: https://github.com/coreylynch/async-rl (fetched 2026-09-03T02:15:10.215496+00:00, sha 030bd9aae9c1)
- Data as of 2026-08-30T08:39:29.467469+00:00.