{
  "id": 415818,
  "title": "#Submission #NeedHelp",
  "url": "/competitions/tlvmc-parkinsons-freezing-gait-prediction/discussion/415818",
  "author_name": "",
  "post_date": "2023-06-08T07:59:40.317565500Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi!</p>\n<p>I ran out of GPU quota for this week and cannot run and submit my notebook. Submission mechanics seems to require submission.csv based on two files in public test set, which would take forever on CPU (and TPU is not allowed according to the competition code).</p>\n<p>I'm sure there should be some walkaround with Collab notebooks or something. Would greatly appreciate your (timely :)) advise on this matter.</p>\n<p>Cheers! </p>",
  "messages": [
    {
      "id": "2292314",
      "postDate": "06/08/2023 07:59:40",
      "content": "<p>Hi!</p>\n<p>I ran out of GPU quota for this week and cannot run and submit my notebook. Submission mechanics seems to require submission.csv based on two files in public test set, which would take forever on CPU (and TPU is not allowed according to the competition code).</p>\n<p>I'm sure there should be some walkaround with Collab notebooks or something. Would greatly appreciate your (timely :)) advise on this matter.</p>\n<p>Cheers! </p>",
      "rawMarkdown": "Hi!\n\nI ran out of GPU quota for this week and cannot run and submit my notebook. Submission mechanics seems to require submission.csv based on two files in public test set, which would take forever on CPU (and TPU is not allowed according to the competition code).\n\nI'm sure there should be some walkaround with Collab notebooks or something. Would greatly appreciate your (timely :)) advise on this matter.\n\nCheers!",
      "votes": null
    },
    {
      "id": "2292829",
      "postDate": "06/08/2023 16:43:26",
      "content": "<blockquote>\n  <p>Submission mechanics seems to require submission.csv based on two files in public test set</p>\n</blockquote>\n<p>There are 2 <em>visible</em> files in the test set, but many <em>hidden</em> files are added to those folders after you submit a notebook for scoring.  Make sure that your notebook is capable of handling all hidden test files -- e.g., iterate through all filepaths returned by <code>os.listdir()</code>.  Take a look at some of the publicly available notebooks for examples.</p>\n<blockquote>\n  <p>I'm sure there should be some walkaround with Collab notebooks or something</p>\n</blockquote>\n<p>You can:</p>\n<ol>\n<li>Train a model with Colab or a local GPU</li>\n<li>Upload the model (or its weights) to Kaggle as a private dataset.  </li>\n<li>Load the saved model in your submission notebook.</li>\n</ol>\n<p>However, submission.csv <strong>must</strong> be generated by your model after submitting your notebook for scoring.</p>",
      "rawMarkdown": ">Submission mechanics seems to require submission.csv based on two files in public test set\n\nThere are 2 *visible* files in the test set, but many *hidden* files are added to those folders after you submit a notebook for scoring.  Make sure that your notebook is capable of handling all hidden test files -- e.g., iterate through all filepaths returned by `os.listdir()`.  Take a look at some of the publicly available notebooks for examples.\n\n>I'm sure there should be some walkaround with Collab notebooks or something\n\nYou can:\n1.  Train a model with Colab or a local GPU\n2.  Upload the model (or its weights) to Kaggle as a private dataset.  \n3.  Load the saved model in your submission notebook.\n\nHowever, submission.csv **must** be generated by your model after submitting your notebook for scoring.",
      "votes": null
    },
    {
      "id": "2292866",
      "postDate": "06/08/2023 17:33:55",
      "content": "<p>Thanks! I wasn't aware that we could upload pretrained models - I thought training should be completed on in Kaggle notebook within a 9h limit as well. That's a major difference - will try to make something out of it in the 7 hours left :)</p>",
      "rawMarkdown": "Thanks! I wasn't aware that we could upload pretrained models - I thought training should be completed on in Kaggle notebook within a 9h limit as well. That's a major difference - will try to make something out of it in the 7 hours left :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2292829,
      "author_name": "quillen",
      "author_url": "",
      "post_date": "06/08/2023 16:43:26",
      "content": "<blockquote>\n  <p>Submission mechanics seems to require submission.csv based on two files in public test set</p>\n</blockquote>\n<p>There are 2 <em>visible</em> files in the test set, but many <em>hidden</em> files are added to those folders after you submit a notebook for scoring.  Make sure that your notebook is capable of handling all hidden test files -- e.g., iterate through all filepaths returned by <code>os.listdir()</code>.  Take a look at some of the publicly available notebooks for examples.</p>\n<blockquote>\n  <p>I'm sure there should be some walkaround with Collab notebooks or something</p>\n</blockquote>\n<p>You can:</p>\n<ol>\n<li>Train a model with Colab or a local GPU</li>\n<li>Upload the model (or its weights) to Kaggle as a private dataset.  </li>\n<li>Load the saved model in your submission notebook.</li>\n</ol>\n<p>However, submission.csv <strong>must</strong> be generated by your model after submitting your notebook for scoring.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2292866,
          "author_name": "victorshlepov",
          "author_url": "",
          "post_date": "06/08/2023 17:33:55",
          "content": "<p>Thanks! I wasn't aware that we could upload pretrained models - I thought training should be completed on in Kaggle notebook within a 9h limit as well. That's a major difference - will try to make something out of it in the 7 hours left :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2292314": "Hi!\n\nI ran out of GPU quota for this week and cannot run and submit my notebook. Submission mechanics seems to require submission.csv based on two files in public test set, which would take forever on CPU (and TPU is not allowed according to the competition code).\n\nI'm sure there should be some walkaround with Collab notebooks or something. Would greatly appreciate your (timely :)) advise on this matter.\n\nCheers!",
    "2292829": ">Submission mechanics seems to require submission.csv based on two files in public test set\n\nThere are 2 *visible* files in the test set, but many *hidden* files are added to those folders after you submit a notebook for scoring.  Make sure that your notebook is capable of handling all hidden test files -- e.g., iterate through all filepaths returned by `os.listdir()`.  Take a look at some of the publicly available notebooks for examples.\n\n>I'm sure there should be some walkaround with Collab notebooks or something\n\nYou can:\n1.  Train a model with Colab or a local GPU\n2.  Upload the model (or its weights) to Kaggle as a private dataset.  \n3.  Load the saved model in your submission notebook.\n\nHowever, submission.csv **must** be generated by your model after submitting your notebook for scoring.",
    "2292866": "Thanks! I wasn't aware that we could upload pretrained models - I thought training should be completed on in Kaggle notebook within a 9h limit as well. That's a major difference - will try to make something out of it in the 7 hours left :)"
  },
  "source": "meta"
}