{
  "id": 177157,
  "title": "Training is not required in Notebooks?",
  "url": "/competitions/lyft-motion-prediction-autonomous-vehicles/discussion/177157",
  "author_name": "",
  "post_date": "2020-08-25T04:08:48.844958500Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello fellows,</p>\n<p>In the Code Requirements section it says \"Please note that for this competition training is not required in Notebooks\". What does that mean? Does that mean we can pre-train a model in our own computer and upload it as external data?</p>\n<p>Another question that is probably (also) silly: In the Data Description there is a statement that \"For your convenience, Lyft's l5kit module is provided via a utility script called kaggle_l5kit. Just attach it to your kernel, and the latest version of l5kit and all dependencies will be available.\" How could we \"attach\" it to our kernel? Since we do not have access to internet, we cannot install l5kit in our kernel right?</p>\n<p>Many thanks!<br>\nFrank</p>",
  "messages": [
    {
      "id": "984359",
      "postDate": "08/25/2020 04:08:48",
      "content": "<p>Hello fellows,</p>\n<p>In the Code Requirements section it says \"Please note that for this competition training is not required in Notebooks\". What does that mean? Does that mean we can pre-train a model in our own computer and upload it as external data?</p>\n<p>Another question that is probably (also) silly: In the Data Description there is a statement that \"For your convenience, Lyft's l5kit module is provided via a utility script called kaggle_l5kit. Just attach it to your kernel, and the latest version of l5kit and all dependencies will be available.\" How could we \"attach\" it to our kernel? Since we do not have access to internet, we cannot install l5kit in our kernel right?</p>\n<p>Many thanks!<br>\nFrank</p>",
      "rawMarkdown": "Hello fellows,\n\nIn the Code Requirements section it says \"Please note that for this competition training is not required in Notebooks\". What does that mean? Does that mean we can pre-train a model in our own computer and upload it as external data?\n\nAnother question that is probably (also) silly: In the Data Description there is a statement that \"For your convenience, Lyft's l5kit module is provided via a utility script called kaggle_l5kit. Just attach it to your kernel, and the latest version of l5kit and all dependencies will be available.\" How could we \"attach\" it to our kernel? Since we do not have access to internet, we cannot install l5kit in our kernel right?\n\nMany thanks!\nFrank",
      "votes": null
    },
    {
      "id": "984427",
      "postDate": "08/25/2020 05:08:02",
      "content": "<p>1) That's correct. You can train on your computer or something like Google Colab then upload your weights as a Kaggle dataset.</p>\n<p>2) I haven't used it yet but if you go to this notebook it looks like you just have to install it at the start like so</p>\n<pre><code>import os\nos.system('pip uninstall typing -y')\nos.system('pip install --ignore-installed --target=/kaggle/working l5kit')\n</code></pre>",
      "rawMarkdown": "1) That's correct. You can train on your computer or something like Google Colab then upload your weights as a Kaggle dataset.\n\n2) I haven't used it yet but if you go to this notebook it looks like you just have to install it at the start like so\n```\nimport os\nos.system('pip uninstall typing -y')\nos.system('pip install --ignore-installed --target=/kaggle/working l5kit')\n```",
      "votes": null
    },
    {
      "id": "984551",
      "postDate": "08/25/2020 06:30:15",
      "content": "<p>Thank you Matthew. That is helpful.</p>",
      "rawMarkdown": "Thank you Matthew. That is helpful.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 984427,
      "author_name": "matthewmasters",
      "author_url": "",
      "post_date": "08/25/2020 05:08:02",
      "content": "<p>1) That's correct. You can train on your computer or something like Google Colab then upload your weights as a Kaggle dataset.</p>\n<p>2) I haven't used it yet but if you go to this notebook it looks like you just have to install it at the start like so</p>\n<pre><code>import os\nos.system('pip uninstall typing -y')\nos.system('pip install --ignore-installed --target=/kaggle/working l5kit')\n</code></pre>",
      "votes": null,
      "replies": [
        {
          "id": 984551,
          "author_name": "frankpanxj",
          "author_url": "",
          "post_date": "08/25/2020 06:30:15",
          "content": "<p>Thank you Matthew. That is helpful.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "984359": "Hello fellows,\n\nIn the Code Requirements section it says \"Please note that for this competition training is not required in Notebooks\". What does that mean? Does that mean we can pre-train a model in our own computer and upload it as external data?\n\nAnother question that is probably (also) silly: In the Data Description there is a statement that \"For your convenience, Lyft's l5kit module is provided via a utility script called kaggle_l5kit. Just attach it to your kernel, and the latest version of l5kit and all dependencies will be available.\" How could we \"attach\" it to our kernel? Since we do not have access to internet, we cannot install l5kit in our kernel right?\n\nMany thanks!\nFrank",
    "984427": "1) That's correct. You can train on your computer or something like Google Colab then upload your weights as a Kaggle dataset.\n\n2) I haven't used it yet but if you go to this notebook it looks like you just have to install it at the start like so\n```\nimport os\nos.system('pip uninstall typing -y')\nos.system('pip install --ignore-installed --target=/kaggle/working l5kit')\n```",
    "984551": "Thank you Matthew. That is helpful."
  },
  "source": "meta"
}