Skip to content

Fail to generate tasks #1

Description

@mykuro

Hi there,

Thanks for sharing this repository. I am interested in your work on EnactToM benchmark and am looking forward to evaluate on it.

I noticed your page mentions "Dataset(coming soon)", but it appears the last commit of the code was about 2 months ago. Thus, I tried to generate tasks as the instruction ./enacttom/run.sh generate --num-tasks 1 --difficulty standard with the default model but failed. Could you release the dataset or tell me the reason why it always couldn't generate tasks the meet the standards. I found that "failure_explanations": ["vase_0 is not is_on_top table_32", "glass_4 is not is_on_top table_32"] even if {"agent": "agent_0","action": "Place[glass_4, on, table_32, None, None]", "success": true,"observation": "Successful execution!\n\nagent_0_VisualSummary:\n- No notable new visual observations"} and {"agent": "agent_2","action": "Place[vase_0, on, table_32, None, None]", "success": true,"observation": "Successful execution!\n\nagent_0_VisualSummary:\n- No notable new visual observations"}

Thank you for your time and contribution!

Image

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