Skip to content

Run on SWEBench need to Edit source code of SWEBench #4

Description

@XYZboom

Currently, we need to edit SWEBench enable to evaluate ourself.
Open swebench/harness/docker_build.py, found this:

eval_commands += [
        reset_tests_command,
        apply_test_patch_command,
        f": '{START_TEST_OUTPUT}'",
        test_command,
        f": '{END_TEST_OUTPUT}'",
        reset_tests_command,  # Revert tests after done, leave the repo in the same state as before
    ]
    return eval_commands

Edit into:

eval_commands += [
        reset_tests_command,
        apply_test_patch_command,
        f"echo '{START_TEST_OUTPUT}'",
        test_command,
        f"echo '{END_TEST_OUTPUT}'",
        reset_tests_command,  # Revert tests after done, leave the repo in the same state as before
    ]
    return eval_commands

Reinstall swebench.
We need to find a way not to edit source code of swebench

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions