Skip to content

[Reliability] Improved Error Parsing for Subprocess Training #17

Description

@DsThakurRawat

Description

The LerobotBackend invokes lerobot-train via a subprocess and currently only reports the exit code on failure.

Rationale

Robot training often fails due to hardware-specific issues (OOM, CUDA errors, disk full). Swallowing these behind a generic 'exited with code 1' message makes debugging difficult for users.

Tasks

  • Implement a regex-based error parser for the subprocess stderr stream.
  • Map common strings (e.g., 'OutOfMemoryError', 'CUDA error: out of memory') to actionable remediation hints.
  • Surface these hints in the FinetuneResult.error payload.

Activity

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

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions