关于GCU、沐曦GPGPU、MLU、0卡V100资源4月7日恢复上架的公告>>> 关于共建具身智能开源数据集的倡议>>> 关于云脑任务中统一路径访问方式的公告>>> 关于将启智集群GPU资源迁移至智算集群的公告>>>
OpenI
/
PARL
mirror of https://gitee.com/paddlepaddle/PARL.git
Not watched
16
5
Fork 9
Code
Releases 9 Wiki Activity
Issues 0
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
501 Commits
126 Branches
63 MiB
1114 Download
 
 
 
 
 
 
Branch: develop
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'develop'
${ noResults }
PARL/examples/DQN_variant
Happy 83a6dcaa3e
dev 2.1.1 (#1009) 		1 year ago
..
.benchmark add torch paddle dqn roadrunner (#695) 2 years ago
README.md dev 2.1.1 (#1009) 1 year ago
atari_agent.py support zero-dim tensor for paddle future version (#986) 1 year ago
atari_model.py fix style (#657) 2 years ago
replay_memory.py add DQN atari examples and DDQN implementations (#624) 2 years ago
requirements.txt dev 2.1.1 (#1009) 1 year ago
train.py dev 2.1.1 (#1009) 1 year ago

README.md

Reproduce DQN with PARL

Based on PARL, the DQN algorithm of deep reinforcement learning has been reproduced, reaching the same level of indicators as the paper in Atari benchmarks.

  • Papers:

DQN in Human-level Control Through Deep Reinforcement Learning

DDQN in Deep Reinforcement Learning with Double Q-learning

Dueling DQN in Dueling Network Architectures for Deep Reinforcement Learning

Atari games introduction

Please see here to know more about Atari games.

Benchmark results

Benchmark results are obtained using different random seeds.

Performance of Dueling DQN on various environments:

result

Performance of Dueling DQN on 55 Atari environments:
Alien (5977) Amidar (364) Assault (9676) Asterix (23800) Asteroids (657)
Atlantis (85633) WizardOfWor (2767) BankHeist (1143) BattleZone (37667) BeamRider (13570)
Berzerk (827) Bowling (47) Boxing (100) Breakout (409) Centipede (5103)
ChopperCommand (1300) CrazyClimber (118733) DemonAttack (167200) DoubleDunk (-1) Enduro (4153)
FishingDerby (-64) Freeway (22) Frostbite (5273) Gopher (11187) Gravitar (0)
Hero (14613) IceHockey (2) Jamesbond (767) Kangaroo (4133) Krull (8856)
KungFuMaster (19933) MontezumaRevenge (0) MsPacman (4013) NameThisGame (10327) Phoenix (7333)
Pitfall (0) Pong (21) PrivateEye (49) Qbert (15275) Riverraid (13410)
RoadRunner (47167) Robotank (27) Seaquest (16573) Skiing (-14409) Solaris (53)
SpaceInvaders (2797) StarGunner (59367) Tennis (0) TimePilot (8200) Tutankham (235)
UpNDown (18153) Venture (0) VideoPinball (745800) YarsRevenge (34346) Zaxxon (13233)

How to use

Dependencies:

Start Training:

# To train a model for Pong game
python train.py

# For more customized arguments
python train.py --help
Delete task

Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.

cancel
confirm

PARL 是一个高性能、灵活的强化学习框架

https://parl.readthedocs.io

Python C++ JavaScript Shell Markdown other

Please read the following content carefully:

Dear OpenI User

Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.

For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》

Agree and continue
Disagree, exit