{
  "id": 473159,
  "title": "Trouble With Submission Error",
  "url": "/competitions/hms-harmful-brain-activity-classification/discussion/473159",
  "author_name": "",
  "post_date": "2024-02-03T18:22:52.409039600Z",
  "votes": 3,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Anyone else having submission error issues?</p>\n<p>I've checked the following:</p>\n<ul>\n<li>The number of input and output rows are the same</li>\n<li>The sum of the output probabilities are<ul>\n<li>Less than or equal to one</li>\n<li>Greater than or equal to zero</li>\n<li>Sum to roughly 1</li></ul></li>\n<li>The output is saved with the only the target columns and <code>eeg_id</code>. No additional index is saved</li>\n<li>The dataloader isn't shuffling the inputs (I don't think)</li>\n</ul>\n<p>Is there anything else to check? Maybe it's ordering? </p>",
  "messages": [
    {
      "id": "2634415",
      "postDate": "02/03/2024 18:22:52",
      "content": "<p>Anyone else having submission error issues?</p>\n<p>I've checked the following:</p>\n<ul>\n<li>The number of input and output rows are the same</li>\n<li>The sum of the output probabilities are<ul>\n<li>Less than or equal to one</li>\n<li>Greater than or equal to zero</li>\n<li>Sum to roughly 1</li></ul></li>\n<li>The output is saved with the only the target columns and <code>eeg_id</code>. No additional index is saved</li>\n<li>The dataloader isn't shuffling the inputs (I don't think)</li>\n</ul>\n<p>Is there anything else to check? Maybe it's ordering? </p>",
      "rawMarkdown": "Anyone else having submission error issues?\n\nI've checked the following:\n- The number of input and output rows are the same\n- The sum of the output probabilities are\n  - Less than or equal to one\n  - Greater than or equal to zero\n  - Sum to roughly 1\n- The output is saved with the only the target columns and `eeg_id`. No additional index is saved\n- The dataloader isn't shuffling the inputs (I don't think)\n\nIs there anything else to check? Maybe it's ordering?",
      "votes": null
    },
    {
      "id": "2634440",
      "postDate": "02/03/2024 18:50:09",
      "content": "<p>What is the error message? How did you created the submission file? I could be the ordering.</p>",
      "rawMarkdown": "What is the error message? How did you created the submission file? I could be the ordering.",
      "votes": null
    },
    {
      "id": "2634447",
      "postDate": "02/03/2024 18:55:38",
      "content": "<p>Standard <code>pd.DataFrame.to_csv('submission.csv', index=False)</code> where I didn't set the index to anything meaningful. The error message is the bland:</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F798021%2Fccf8ed3a9ab8fc54ca5ba176e88975c1%2FScreenshot%20from%202024-02-03%2010-54-24.png?generation=1706986482881566&amp;alt=media\"></p>\n<p>I'm going to try making double sure there's no NaNs or rounding errors (I have a sneaking suspicion that the writer is introducing a small rounding error).</p>",
      "rawMarkdown": "Standard `pd.DataFrame.to_csv('submission.csv', index=False)` where I didn't set the index to anything meaningful. The error message is the bland:\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F798021%2Fccf8ed3a9ab8fc54ca5ba176e88975c1%2FScreenshot%20from%202024-02-03%2010-54-24.png?generation=1706986482881566&alt=media)\n\nI'm going to try making double sure there's no NaNs or rounding errors (I have a sneaking suspicion that the writer is introducing a small rounding error).",
      "votes": null
    },
    {
      "id": "2635679",
      "postDate": "02/04/2024 14:54:56",
      "content": "<p>The competition uses hidden test data as well to score the submissions. So, the issue maybe that your code is not handling test data of different shapes and values very well. This <a href=\"https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/473373\" target=\"_blank\">discussion</a> maybe useful</p>",
      "rawMarkdown": "The competition uses hidden test data as well to score the submissions. So, the issue maybe that your code is not handling test data of different shapes and values very well. This [discussion](https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/473373) maybe useful",
      "votes": null
    },
    {
      "id": "2637010",
      "postDate": "02/05/2024 12:40:31",
      "content": "<p>Fixed it! Diagnosed by using the training dataset and for some weird reason the model was returning NaNs for some rows. The fix was a <code>fillna</code> for now, but I'm looking closer into how to prevent softmax from returning NaN.</p>",
      "rawMarkdown": "Fixed it! Diagnosed by using the training dataset and for some weird reason the model was returning NaNs for some rows. The fix was a `fillna` for now, but I'm looking closer into how to prevent softmax from returning NaN.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2634440,
      "author_name": "simple11",
      "author_url": "",
      "post_date": "02/03/2024 18:50:09",
      "content": "<p>What is the error message? How did you created the submission file? I could be the ordering.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2634447,
          "author_name": "kwierman",
          "author_url": "",
          "post_date": "02/03/2024 18:55:38",
          "content": "<p>Standard <code>pd.DataFrame.to_csv('submission.csv', index=False)</code> where I didn't set the index to anything meaningful. The error message is the bland:</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F798021%2Fccf8ed3a9ab8fc54ca5ba176e88975c1%2FScreenshot%20from%202024-02-03%2010-54-24.png?generation=1706986482881566&amp;alt=media\"></p>\n<p>I'm going to try making double sure there's no NaNs or rounding errors (I have a sneaking suspicion that the writer is introducing a small rounding error).</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2635679,
      "author_name": "eouedraogo4",
      "author_url": "",
      "post_date": "02/04/2024 14:54:56",
      "content": "<p>The competition uses hidden test data as well to score the submissions. So, the issue maybe that your code is not handling test data of different shapes and values very well. This <a href=\"https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/473373\" target=\"_blank\">discussion</a> maybe useful</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2637010,
      "author_name": "kwierman",
      "author_url": "",
      "post_date": "02/05/2024 12:40:31",
      "content": "<p>Fixed it! Diagnosed by using the training dataset and for some weird reason the model was returning NaNs for some rows. The fix was a <code>fillna</code> for now, but I'm looking closer into how to prevent softmax from returning NaN.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2634415": "Anyone else having submission error issues?\n\nI've checked the following:\n- The number of input and output rows are the same\n- The sum of the output probabilities are\n  - Less than or equal to one\n  - Greater than or equal to zero\n  - Sum to roughly 1\n- The output is saved with the only the target columns and `eeg_id`. No additional index is saved\n- The dataloader isn't shuffling the inputs (I don't think)\n\nIs there anything else to check? Maybe it's ordering?",
    "2634440": "What is the error message? How did you created the submission file? I could be the ordering.",
    "2634447": "Standard `pd.DataFrame.to_csv('submission.csv', index=False)` where I didn't set the index to anything meaningful. The error message is the bland:\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F798021%2Fccf8ed3a9ab8fc54ca5ba176e88975c1%2FScreenshot%20from%202024-02-03%2010-54-24.png?generation=1706986482881566&alt=media)\n\nI'm going to try making double sure there's no NaNs or rounding errors (I have a sneaking suspicion that the writer is introducing a small rounding error).",
    "2635679": "The competition uses hidden test data as well to score the submissions. So, the issue maybe that your code is not handling test data of different shapes and values very well. This [discussion](https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/473373) maybe useful",
    "2637010": "Fixed it! Diagnosed by using the training dataset and for some weird reason the model was returning NaNs for some rows. The fix was a `fillna` for now, but I'm looking closer into how to prevent softmax from returning NaN."
  },
  "source": "meta"
}