Skip to content

feat: add robosuite support for sac#20

Open
OrangeX4 wants to merge 5 commits into
LAMDA-RL:masterfrom
OrangeX4:sac
Open

feat: add robosuite support for sac#20
OrangeX4 wants to merge 5 commits into
LAMDA-RL:masterfrom
OrangeX4:sac

Conversation

@OrangeX4

Copy link
Copy Markdown
python reproduce/sac/run_sac_online.py --config reproduce/sac/config/robosuite/Door-Panda.py

Comment thread reproduce/sac/config/robosuite/base.py Outdated
log_interval = 10
warmup_epoch = 2
random_policy_epoch = 5
max_trajectory_length = 1000

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这里的max_trajectory_length和环境的horizon都设置成500

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

同时环境还要添加controller=OSC_POSE,hard_reset=False这两个kwargs


if i_epoch % args.save_interval == 0:
logger.log_object(name=f"policy_{i_epoch}.pt", object=policy.state_dict(), path=f"./out/sac/{args.name}/{args.env}/seed{args.seed}/policy/") No newline at end of file
logger.log_object(name=f"policy_{i_epoch}.pt", object=policy.state_dict(), path=f"./out/sac/{args.name}/{args.env}/seed{args.seed}/policy/")

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这里的train loop的逻辑要变更,改成:训练N1个epoch,其中每个epoch会先收集N2个episode,然后训练N3个gradient step,其中N1=2000, N2=10, N3=1000

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants