Skip to content

Question for training dreamer_v3 on InvertedPendulum-v4 #330

@anthony0727

Description

@anthony0727

I ran below command to train InvertedPendulum-v4,

MUJOCO_GL=egl python sheeprl.py     exp=dreamer_v3     env=mujoco     env.id=InvertedPendulum-v4     fabric.accelerator=cuda     fabric.devices=1     fabric.strategy=ddp algo.mlp_keys.encoder=[state] algo.mlp_keys.decoder=[state] algo.cnn_keys.encoder=[] algo.cnn_keys.decoder=[]

but it the agent is not learning (expected to achieve 1000 score)

Rank-0: policy_step=829396, reward_env_1=4.0                                                                                                                                               
Rank-0: policy_step=829400, reward_env_2=4.0                                                                                                                                               
Rank-0: policy_step=829404, reward_env_0=4.0                                                                                                                                               
Rank-0: policy_step=829404, reward_env_3=4.0
Rank-0: policy_step=829412, reward_env_1=4.0
Rank-0: policy_step=829416, reward_env_2=4.0

Am I missing something?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions