{
  "id": 167865,
  "title": "Train locally, infer in notebook?",
  "url": "/competitions/birdsong-recognition/discussion/167865",
  "author_name": "",
  "post_date": "2020-07-18T05:47:29.833724300Z",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello, first time doing a Kaggle competition after quite a while and confused about the Notebook-only rules. Can I train a model locally on my personal machine and later load the model checkpoint and run inference in a (Kaggle) notebook? Does that count as a pretrained model?</p>\n\n<p>The FAQ (<a href=\"https://www.kaggle.com/docs/competitions#notebooks-only-FAQ\">https://www.kaggle.com/docs/competitions#notebooks-only-FAQ</a>) does not explicitly mention this.</p>\n\n<p>Thanks in advance!</p>",
  "messages": [
    {
      "id": "933911",
      "postDate": "07/18/2020 05:47:29",
      "content": "<p>Hello, first time doing a Kaggle competition after quite a while and confused about the Notebook-only rules. Can I train a model locally on my personal machine and later load the model checkpoint and run inference in a (Kaggle) notebook? Does that count as a pretrained model?</p>\n\n<p>The FAQ (<a href=\"https://www.kaggle.com/docs/competitions#notebooks-only-FAQ\">https://www.kaggle.com/docs/competitions#notebooks-only-FAQ</a>) does not explicitly mention this.</p>\n\n<p>Thanks in advance!</p>",
      "rawMarkdown": "Hello, first time doing a Kaggle competition after quite a while and confused about the Notebook-only rules. Can I train a model locally on my personal machine and later load the model checkpoint and run inference in a (Kaggle) notebook? Does that count as a pretrained model?\n\nThe FAQ (https://www.kaggle.com/docs/competitions#notebooks-only-FAQ) does not explicitly mention this.\n\nThanks in advance!",
      "votes": null
    },
    {
      "id": "933915",
      "postDate": "07/18/2020 05:53:22",
      "content": "<p><a href=\"/danielhavir\">@danielhavir</a> training locally and loading the checkpoint in inference notebook is totally valid strategy. It does not not count as pretrained model and you DON'T have any obligation to share the weights publicly.  although while training locally if you use any pretrained model (eg resnet trained on imagenet) you should disclose it the the external data thread. I hope this clears your doubts and have a nice kaggle journey.</p>",
      "rawMarkdown": "danielhavir training locally and loading the checkpoint in inference notebook is totally valid strategy. It does not not count as pretrained model and you DON'T have any obligation to share the weights publicly.  although while training locally if you use any pretrained model (eg resnet trained on imagenet) you should disclose it the the external data thread. I hope this clears your doubts and have a nice kaggle journey.",
      "votes": null
    },
    {
      "id": "933918",
      "postDate": "07/18/2020 06:00:38",
      "content": "<p>Thank you Dhananjay!</p>",
      "rawMarkdown": "Thank you Dhananjay!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 933915,
      "author_name": "dhananjay3",
      "author_url": "",
      "post_date": "07/18/2020 05:53:22",
      "content": "<p><a href=\"/danielhavir\">@danielhavir</a> training locally and loading the checkpoint in inference notebook is totally valid strategy. It does not not count as pretrained model and you DON'T have any obligation to share the weights publicly.  although while training locally if you use any pretrained model (eg resnet trained on imagenet) you should disclose it the the external data thread. I hope this clears your doubts and have a nice kaggle journey.</p>",
      "votes": null,
      "replies": [
        {
          "id": 933918,
          "author_name": "danielhavir",
          "author_url": "",
          "post_date": "07/18/2020 06:00:38",
          "content": "<p>Thank you Dhananjay!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "933911": "Hello, first time doing a Kaggle competition after quite a while and confused about the Notebook-only rules. Can I train a model locally on my personal machine and later load the model checkpoint and run inference in a (Kaggle) notebook? Does that count as a pretrained model?\n\nThe FAQ (https://www.kaggle.com/docs/competitions#notebooks-only-FAQ) does not explicitly mention this.\n\nThanks in advance!",
    "933915": "danielhavir training locally and loading the checkpoint in inference notebook is totally valid strategy. It does not not count as pretrained model and you DON'T have any obligation to share the weights publicly.  although while training locally if you use any pretrained model (eg resnet trained on imagenet) you should disclose it the the external data thread. I hope this clears your doubts and have a nice kaggle journey.",
    "933918": "Thank you Dhananjay!"
  },
  "source": "meta"
}