{
  "id": 545495,
  "title": "Notebook failing during submission",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/545495",
  "author_name": "",
  "post_date": "2024-11-10T15:56:22.496644300Z",
  "votes": 1,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I am trying to submit my notebook but cannot due to an unknown error. When I commit the notebook, it runs successfully. However, when I submit it, I get a Notebook Inference Server Error, but no details. How do I fix this?</p>",
  "messages": [
    {
      "id": "3041624",
      "postDate": "11/10/2024 15:56:22",
      "content": "<p>I am trying to submit my notebook but cannot due to an unknown error. When I commit the notebook, it runs successfully. However, when I submit it, I get a Notebook Inference Server Error, but no details. How do I fix this?</p>",
      "rawMarkdown": "I am trying to submit my notebook but cannot due to an unknown error. When I commit the notebook, it runs successfully. However, when I submit it, I get a Notebook Inference Server Error, but no details. How do I fix this?",
      "votes": null
    },
    {
      "id": "3043089",
      "postDate": "11/12/2024 05:16:32",
      "content": "<p>check you can handle missing symbols and new symbols and that you aren't changing polars dataframe dtypes somehow (e.g. during infilling Null value etc)</p>",
      "rawMarkdown": "check you can handle missing symbols and new symbols and that you aren't changing polars dataframe dtypes somehow (e.g. during infilling Null value etc)",
      "votes": null
    },
    {
      "id": "3043915",
      "postDate": "11/12/2024 19:53:25",
      "content": "<p>what do you mean by missing (rather than just new) symbols, assuming you're refering to symbol_id's?</p>",
      "rawMarkdown": "what do you mean by missing (rather than just new) symbols, assuming you're refering to symbol_id's?",
      "votes": null
    },
    {
      "id": "3043944",
      "postDate": "11/12/2024 20:27:02",
      "content": "<p>But if my code is properly written then it would work the same in testing and production, right?</p>",
      "rawMarkdown": "But if my code is properly written then it would work the same in testing and production, right?",
      "votes": null
    },
    {
      "id": "3044020",
      "postDate": "11/12/2024 23:01:49",
      "content": "<p>yes. You might only get 30 symbol_id one day instead of the full range for example</p>",
      "rawMarkdown": "yes. You might only get 30 symbol_id one day instead of the full range for example",
      "votes": null
    },
    {
      "id": "3044022",
      "postDate": "11/12/2024 23:03:05",
      "content": "<p>The testing sample does not cover many edge cases. I suggest you write a local implementation of the eval server that serves you historical training data in the same manner. Run it over 5-10 days of historical data and see if you have issues arise. I think someone has already shared a possible implementation for this in a notebook.</p>",
      "rawMarkdown": "The testing sample does not cover many edge cases. I suggest you write a local implementation of the eval server that serves you historical training data in the same manner. Run it over 5-10 days of historical data and see if you have issues arise. I think someone has already shared a possible implementation for this in a notebook.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3043089,
      "author_name": "michaeltimbs",
      "author_url": "",
      "post_date": "11/12/2024 05:16:32",
      "content": "<p>check you can handle missing symbols and new symbols and that you aren't changing polars dataframe dtypes somehow (e.g. during infilling Null value etc)</p>",
      "votes": null,
      "replies": [
        {
          "id": 3043915,
          "author_name": "boubous",
          "author_url": "",
          "post_date": "11/12/2024 19:53:25",
          "content": "<p>what do you mean by missing (rather than just new) symbols, assuming you're refering to symbol_id's?</p>",
          "votes": null,
          "replies": [
            {
              "id": 3044020,
              "author_name": "michaeltimbs",
              "author_url": "",
              "post_date": "11/12/2024 23:01:49",
              "content": "<p>yes. You might only get 30 symbol_id one day instead of the full range for example</p>",
              "votes": null,
              "replies": []
            }
          ]
        },
        {
          "id": 3043944,
          "author_name": "samirrangwalla",
          "author_url": "",
          "post_date": "11/12/2024 20:27:02",
          "content": "<p>But if my code is properly written then it would work the same in testing and production, right?</p>",
          "votes": null,
          "replies": [
            {
              "id": 3044022,
              "author_name": "michaeltimbs",
              "author_url": "",
              "post_date": "11/12/2024 23:03:05",
              "content": "<p>The testing sample does not cover many edge cases. I suggest you write a local implementation of the eval server that serves you historical training data in the same manner. Run it over 5-10 days of historical data and see if you have issues arise. I think someone has already shared a possible implementation for this in a notebook.</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3041624": "I am trying to submit my notebook but cannot due to an unknown error. When I commit the notebook, it runs successfully. However, when I submit it, I get a Notebook Inference Server Error, but no details. How do I fix this?",
    "3043089": "check you can handle missing symbols and new symbols and that you aren't changing polars dataframe dtypes somehow (e.g. during infilling Null value etc)",
    "3043915": "what do you mean by missing (rather than just new) symbols, assuming you're refering to symbol_id's?",
    "3043944": "But if my code is properly written then it would work the same in testing and production, right?",
    "3044020": "yes. You might only get 30 symbol_id one day instead of the full range for example",
    "3044022": "The testing sample does not cover many edge cases. I suggest you write a local implementation of the eval server that serves you historical training data in the same manner. Run it over 5-10 days of historical data and see if you have issues arise. I think someone has already shared a possible implementation for this in a notebook."
  },
  "source": "meta"
}