{
  "id": 593927,
  "title": "A notebook version successfully ran, but during submission rerun, the sub failed within one minute",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/593927",
  "author_name": "hsiaosuan",
  "post_date": "2025-07-31T12:06:31.414000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<h2>Notebook successfully ran.</h2>\n<p>My notebook runs successfully during the initial <strong>pre-run</strong>, including full model loading, feature engineering, and prediction. However, when I submit it to the leaderboard (triggering a rerun), it immediately fails with the following message: <br>\n<code>Inference server never started</code></p>\n<p>Other people have encountered this problem because they run the notebook <strong>too long</strong> before they call the inference server. However, my submission crash happened within one minute. Is there some other hidden contraints I am not aware? Such as a hidden time limite per sample sequence (which limits feature engineering).</p>",
  "messages": [
    {
      "id": 3258887,
      "postDate": "2025-07-31T12:06:31.413Z",
      "content": "<h2>Notebook successfully ran.</h2>\n<p>My notebook runs successfully during the initial <strong>pre-run</strong>, including full model loading, feature engineering, and prediction. However, when I submit it to the leaderboard (triggering a rerun), it immediately fails with the following message: <br>\n<code>Inference server never started</code></p>\n<p>Other people have encountered this problem because they run the notebook <strong>too long</strong> before they call the inference server. However, my submission crash happened within one minute. Is there some other hidden contraints I am not aware? Such as a hidden time limite per sample sequence (which limits feature engineering).</p>",
      "rawMarkdown": "## Notebook successfully ran. \nMy notebook runs successfully during the initial **pre-run**, including full model loading, feature engineering, and prediction. However, when I submit it to the leaderboard (triggering a rerun), it immediately fails with the following message: \n`Inference server never started`\n\nOther people have encountered this problem because they run the notebook **too long** before they call the inference server. However, my submission crash happened within one minute. Is there some other hidden contraints I am not aware? Such as a hidden time limite per sample sequence (which limits feature engineering).\n\n"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3258887": "## Notebook successfully ran. \nMy notebook runs successfully during the initial **pre-run**, including full model loading, feature engineering, and prediction. However, when I submit it to the leaderboard (triggering a rerun), it immediately fails with the following message: \n`Inference server never started`\n\nOther people have encountered this problem because they run the notebook **too long** before they call the inference server. However, my submission crash happened within one minute. Is there some other hidden contraints I am not aware? Such as a hidden time limite per sample sequence (which limits feature engineering).\n\n"
  }
}