Robotics

Clarification on Docker Image Size & Multi-Environment Setup (Local Pass, Submission Fails)

56GB Docker image and multi-process setup cause submission failures despite local success...

Deep Dive

In the AI for Industry Challenge, user deepkevin0122 posted a clarification regarding submission failures after local testing passes. Their Docker image is roughly 56GB, raising concerns about implicit or explicit size limits. They also describe a multi-environment setup: a ROS 2 lifecycle node (aic_model) runs in one environment while model inference runs in a separate environment launched via a subprocess. This configuration works locally but fails after submission.

The participant seeks guidance on whether Docker image size could cause upload, deployment, or evaluation failures, and whether multi-process or multi-Conda/Micromamba environment setups are allowed. The challenge platform's rules on these technical details are not explicitly documented, leading to confusion. Understanding these constraints is essential for participants to debug their solutions and ensure successful evaluation.

Key Points
  • Docker image size is ~56GB, potentially exceeding upload or evaluation limits
  • Multi-environment setup uses ROS 2 lifecycle node and separate inference subprocess
  • Local testing passes but submission fails; unclear if multi-process is allowed

Why It Matters

Docker constraints and multi-environment rules are key to debugging failed submissions in AI competitions.