{
  "id": 386387,
  "title": "Offline Notebooks - Installing Packages",
  "url": "/competitions/nfl-player-contact-detection/discussion/386387",
  "author_name": "",
  "post_date": "2023-02-12T19:57:55.895548Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Are we allowed to have the submission notebooks be connected to the internet to download the respective packages, then turn them offline for submission (File Persistence)? Or is the expectation that everything we do on it should be offline?</p>",
  "messages": [
    {
      "id": "2141496",
      "postDate": "02/12/2023 19:57:55",
      "content": "<p>Are we allowed to have the submission notebooks be connected to the internet to download the respective packages, then turn them offline for submission (File Persistence)? Or is the expectation that everything we do on it should be offline?</p>",
      "rawMarkdown": "Are we allowed to have the submission notebooks be connected to the internet to download the respective packages, then turn them offline for submission (File Persistence)? Or is the expectation that everything we do on it should be offline?",
      "votes": null
    },
    {
      "id": "2141547",
      "postDate": "02/12/2023 20:51:03",
      "content": "<p>Everything in the submission has to be done offline. You can install packages by first (outside of the submission), creating a private dataset with a .whl file of the package, and then install it with:</p>\n<p>!pip install ../input/your-dataset-name/package-name.whl</p>\n<p>That way, your submission will be entirely offline.</p>",
      "rawMarkdown": "Everything in the submission has to be done offline. You can install packages by first (outside of the submission), creating a private dataset with a .whl file of the package, and then install it with:\n\n!pip install ../input/your-dataset-name/package-name.whl\n\nThat way, your submission will be entirely offline.",
      "votes": null
    },
    {
      "id": "2141558",
      "postDate": "02/12/2023 21:03:20",
      "content": "<p>Thanks! Wish there was a way for the package management to be done a bit easier. For example, to work with video_reader from torchvision, it has to be built from source. Any suggestions on how to do this?</p>",
      "rawMarkdown": "Thanks! Wish there was a way for the package management to be done a bit easier. For example, to work with video_reader from torchvision, it has to be built from source. Any suggestions on how to do this?",
      "votes": null
    },
    {
      "id": "2141581",
      "postDate": "02/12/2023 21:53:30",
      "content": "<p>Hm, not sure. Could you build it locally and upload just the binary? You can probably get the OS, etc from a running kaggle kernel, and build it for that?</p>",
      "rawMarkdown": "Hm, not sure. Could you build it locally and upload just the binary? You can probably get the OS, etc from a running kaggle kernel, and build it for that?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2141547,
      "author_name": "chris62",
      "author_url": "",
      "post_date": "02/12/2023 20:51:03",
      "content": "<p>Everything in the submission has to be done offline. You can install packages by first (outside of the submission), creating a private dataset with a .whl file of the package, and then install it with:</p>\n<p>!pip install ../input/your-dataset-name/package-name.whl</p>\n<p>That way, your submission will be entirely offline.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2141558,
          "author_name": "reubencpereira",
          "author_url": "",
          "post_date": "02/12/2023 21:03:20",
          "content": "<p>Thanks! Wish there was a way for the package management to be done a bit easier. For example, to work with video_reader from torchvision, it has to be built from source. Any suggestions on how to do this?</p>",
          "votes": null,
          "replies": [
            {
              "id": 2141581,
              "author_name": "chris62",
              "author_url": "",
              "post_date": "02/12/2023 21:53:30",
              "content": "<p>Hm, not sure. Could you build it locally and upload just the binary? You can probably get the OS, etc from a running kaggle kernel, and build it for that?</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2141496": "Are we allowed to have the submission notebooks be connected to the internet to download the respective packages, then turn them offline for submission (File Persistence)? Or is the expectation that everything we do on it should be offline?",
    "2141547": "Everything in the submission has to be done offline. You can install packages by first (outside of the submission), creating a private dataset with a .whl file of the package, and then install it with:\n\n!pip install ../input/your-dataset-name/package-name.whl\n\nThat way, your submission will be entirely offline.",
    "2141558": "Thanks! Wish there was a way for the package management to be done a bit easier. For example, to work with video_reader from torchvision, it has to be built from source. Any suggestions on how to do this?",
    "2141581": "Hm, not sure. Could you build it locally and upload just the binary? You can probably get the OS, etc from a running kaggle kernel, and build it for that?"
  },
  "source": "meta"
}