{
  "id": 98593,
  "title": "Any help on Submission Error?",
  "url": "/competitions/aptos2019-blindness-detection/discussion/98593",
  "author_name": "",
  "post_date": "2019-07-04T20:03:19.511195400Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi all,</p>\n\n<p>Even my slimmed down, prediction only kernel (<a href=\"https://www.kaggle.com/belbert/fork-of-aptos-2019-blindness-detection?scriptVersionId=16735464\">APTOS Kernel</a>) continues to have a submission error, but I cannot for the life of me figure out why.  I have continued to pare it back until it is just the prediction part, and each time it runs without any error in the kernel or commit, but has a submission error after I submit it.</p>\n\n<p>Can anyone help debug this?  I'm at a loss for what could be going on here and don't want to continue improving my algorithm until I can at least submit consistently.</p>\n\n<p>Thanks,\nBen</p>",
  "messages": [
    {
      "id": "568378",
      "postDate": "07/04/2019 20:03:19",
      "content": "<p>Hi all,</p>\n\n<p>Even my slimmed down, prediction only kernel (<a href=\"https://www.kaggle.com/belbert/fork-of-aptos-2019-blindness-detection?scriptVersionId=16735464\">APTOS Kernel</a>) continues to have a submission error, but I cannot for the life of me figure out why.  I have continued to pare it back until it is just the prediction part, and each time it runs without any error in the kernel or commit, but has a submission error after I submit it.</p>\n\n<p>Can anyone help debug this?  I'm at a loss for what could be going on here and don't want to continue improving my algorithm until I can at least submit consistently.</p>\n\n<p>Thanks,\nBen</p>",
      "rawMarkdown": "Hi all,\n\nEven my slimmed down, prediction only kernel ([APTOS Kernel](https://www.kaggle.com/belbert/fork-of-aptos-2019-blindness-detection?scriptVersionId=16735464)) continues to have a submission error, but I cannot for the life of me figure out why.  I have continued to pare it back until it is just the prediction part, and each time it runs without any error in the kernel or commit, but has a submission error after I submit it.\n\nCan anyone help debug this?  I'm at a loss for what could be going on here and don't want to continue improving my algorithm until I can at least submit consistently.\n\nThanks,\nBen",
      "votes": null
    },
    {
      "id": "568397",
      "postDate": "07/04/2019 21:04:08",
      "content": "<p>I have a lot of problem with submissions. \nThe following idea helped me. There are two datasets private and public. Your submission should have exacly the same ID set as the sample_submition file. In this case, your solution will be successfully executed on the private. </p>",
      "rawMarkdown": "I have a lot of problem with submissions. \nThe following idea helped me. There are two datasets private and public. Your submission should have exacly the same ID set as the sample_submition file. In this case, your solution will be successfully executed on the private.",
      "votes": null
    },
    {
      "id": "568404",
      "postDate": "07/04/2019 21:17:49",
      "content": "<p>Thanks for the tip, Alexander!  I am going to try this now--I had previously read the \"Problems with Submissions Solutions\" thread wrong (I read as you <strong>shouldn't</strong> use the sample_submission.csv 🙄 </p>",
      "rawMarkdown": "Thanks for the tip, Alexander!  I am going to try this now--I had previously read the \"Problems with Submissions Solutions\" thread wrong (I read as you **shouldn't** use the sample_submission.csv 🙄",
      "votes": null
    },
    {
      "id": "568747",
      "postDate": "07/05/2019 11:29:50",
      "content": "<p>I too faced the same problem. In my case, I had restricted (hardcoded) the number of predictions in submissions CSV to be equal to the number of entries in the public dataset alone, though I had processed all entries in public + unseen test dataset. The submission CSV was created with ID column having all entries in public + unseen test dataset, whereas the predicted diagnosis column was not populated for all. Because of this mismatch it resulted in error. It was fixed by dynamically calculating the number of predictions and populated to submissions CSV.</p>",
      "rawMarkdown": "I too faced the same problem. In my case, I had restricted (hardcoded) the number of predictions in submissions CSV to be equal to the number of entries in the public dataset alone, though I had processed all entries in public + unseen test dataset. The submission CSV was created with ID column having all entries in public + unseen test dataset, whereas the predicted diagnosis column was not populated for all. Because of this mismatch it resulted in error. It was fixed by dynamically calculating the number of predictions and populated to submissions CSV.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 568397,
      "author_name": "antilmb",
      "author_url": "",
      "post_date": "07/04/2019 21:04:08",
      "content": "<p>I have a lot of problem with submissions. \nThe following idea helped me. There are two datasets private and public. Your submission should have exacly the same ID set as the sample_submition file. In this case, your solution will be successfully executed on the private. </p>",
      "votes": null,
      "replies": [
        {
          "id": 568404,
          "author_name": "belbert",
          "author_url": "",
          "post_date": "07/04/2019 21:17:49",
          "content": "<p>Thanks for the tip, Alexander!  I am going to try this now--I had previously read the \"Problems with Submissions Solutions\" thread wrong (I read as you <strong>shouldn't</strong> use the sample_submission.csv 🙄 </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 568747,
      "author_name": "sornavel",
      "author_url": "",
      "post_date": "07/05/2019 11:29:50",
      "content": "<p>I too faced the same problem. In my case, I had restricted (hardcoded) the number of predictions in submissions CSV to be equal to the number of entries in the public dataset alone, though I had processed all entries in public + unseen test dataset. The submission CSV was created with ID column having all entries in public + unseen test dataset, whereas the predicted diagnosis column was not populated for all. Because of this mismatch it resulted in error. It was fixed by dynamically calculating the number of predictions and populated to submissions CSV.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "568378": "Hi all,\n\nEven my slimmed down, prediction only kernel ([APTOS Kernel](https://www.kaggle.com/belbert/fork-of-aptos-2019-blindness-detection?scriptVersionId=16735464)) continues to have a submission error, but I cannot for the life of me figure out why.  I have continued to pare it back until it is just the prediction part, and each time it runs without any error in the kernel or commit, but has a submission error after I submit it.\n\nCan anyone help debug this?  I'm at a loss for what could be going on here and don't want to continue improving my algorithm until I can at least submit consistently.\n\nThanks,\nBen",
    "568397": "I have a lot of problem with submissions. \nThe following idea helped me. There are two datasets private and public. Your submission should have exacly the same ID set as the sample_submition file. In this case, your solution will be successfully executed on the private.",
    "568404": "Thanks for the tip, Alexander!  I am going to try this now--I had previously read the \"Problems with Submissions Solutions\" thread wrong (I read as you **shouldn't** use the sample_submission.csv 🙄",
    "568747": "I too faced the same problem. In my case, I had restricted (hardcoded) the number of predictions in submissions CSV to be equal to the number of entries in the public dataset alone, though I had processed all entries in public + unseen test dataset. The submission CSV was created with ID column having all entries in public + unseen test dataset, whereas the predicted diagnosis column was not populated for all. Because of this mismatch it resulted in error. It was fixed by dynamically calculating the number of predictions and populated to submissions CSV."
  },
  "source": "meta"
}