{
  "id": 205596,
  "title": "Submission Failure (running time)",
  "url": "/competitions/riiid-test-answer-prediction/discussion/205596",
  "author_name": "",
  "post_date": "2020-12-20T19:32:28.427908900Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I am new to the competition and spent quite a few time to working on the projects. Does anyone has the issue of submission failure? It took me 9 hours to submit and got an error then. But it runs successfully if I simply just run the notebook. Appreciate anyone may help on it.</p>",
  "messages": [
    {
      "id": "1120380",
      "postDate": "12/20/2020 19:32:28",
      "content": "<p>I am new to the competition and spent quite a few time to working on the projects. Does anyone has the issue of submission failure? It took me 9 hours to submit and got an error then. But it runs successfully if I simply just run the notebook. Appreciate anyone may help on it.</p>",
      "rawMarkdown": "I am new to the competition and spent quite a few time to working on the projects. Does anyone has the issue of submission failure? It took me 9 hours to submit and got an error then. But it runs successfully if I simply just run the notebook. Appreciate anyone may help on it.",
      "votes": null
    },
    {
      "id": "1121908",
      "postDate": "12/22/2020 02:39:22",
      "content": "<p>The main reason of  it's because the test dataset have about 20+ millions of rows, and take a long time to process your work.<br>\nA good performance notebook take around 4 hours to submit successfuly .</p>\n<p>Some Advices:</p>\n<ul>\n<li>Improve the merges / concats speed of your feature engineering (see: <a href=\"https://www.kaggle.com/tkm2261/fast-pandas-left-join-357x-faster-than-pd-merge\" target=\"_blank\">FAST PANDAS LEFT JOIN (357x faster than pd.merge)</a>)</li>\n<li>If your model training take too long, consider training  separately and then load the  pre-trained model on submission.</li>\n</ul>",
      "rawMarkdown": "The main reason of  it's because the test dataset have about 20+ millions of rows, and take a long time to process your work.\nA good performance notebook take around 4 hours to submit successfuly .\n\nSome Advices:\n- Improve the merges / concats speed of your feature engineering (see: [FAST PANDAS LEFT JOIN (357x faster than pd.merge)](https://www.kaggle.com/tkm2261/fast-pandas-left-join-357x-faster-than-pd-merge))\n- If your model training take too long, consider training  separately and then load the  pre-trained model on submission.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1121908,
      "author_name": "eduardorenz",
      "author_url": "",
      "post_date": "12/22/2020 02:39:22",
      "content": "<p>The main reason of  it's because the test dataset have about 20+ millions of rows, and take a long time to process your work.<br>\nA good performance notebook take around 4 hours to submit successfuly .</p>\n<p>Some Advices:</p>\n<ul>\n<li>Improve the merges / concats speed of your feature engineering (see: <a href=\"https://www.kaggle.com/tkm2261/fast-pandas-left-join-357x-faster-than-pd-merge\" target=\"_blank\">FAST PANDAS LEFT JOIN (357x faster than pd.merge)</a>)</li>\n<li>If your model training take too long, consider training  separately and then load the  pre-trained model on submission.</li>\n</ul>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1120380": "I am new to the competition and spent quite a few time to working on the projects. Does anyone has the issue of submission failure? It took me 9 hours to submit and got an error then. But it runs successfully if I simply just run the notebook. Appreciate anyone may help on it.",
    "1121908": "The main reason of  it's because the test dataset have about 20+ millions of rows, and take a long time to process your work.\nA good performance notebook take around 4 hours to submit successfuly .\n\nSome Advices:\n- Improve the merges / concats speed of your feature engineering (see: [FAST PANDAS LEFT JOIN (357x faster than pd.merge)](https://www.kaggle.com/tkm2261/fast-pandas-left-join-357x-faster-than-pd-merge))\n- If your model training take too long, consider training  separately and then load the  pre-trained model on submission."
  },
  "source": "meta"
}