{
  "id": 166468,
  "title": "Are we allowed to train on our own machines?",
  "url": "/competitions/landmark-retrieval-2020/discussion/166468",
  "author_name": "",
  "post_date": "2020-07-13T02:31:26.719349400Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>This dataset is fairly large and this puts a large constraint on the time.  Even pretrained models can't really be trained for more that 7-8 epochs.  Are we allowed to do training on our own machines?</p>",
  "messages": [
    {
      "id": "926801",
      "postDate": "07/13/2020 02:31:26",
      "content": "<p>This dataset is fairly large and this puts a large constraint on the time.  Even pretrained models can't really be trained for more that 7-8 epochs.  Are we allowed to do training on our own machines?</p>",
      "rawMarkdown": "This dataset is fairly large and this puts a large constraint on the time.  Even pretrained models can't really be trained for more that 7-8 epochs.  Are we allowed to do training on our own machines?",
      "votes": null
    },
    {
      "id": "926879",
      "postDate": "07/13/2020 04:17:07",
      "content": "<p>I think you can train the model on your own machine, then upload the model as a dataset on kaggle, and then incorporate that for you submission notebook. I think the main constraint is submitting a model instead of predictions, so if you can sort of follow the sample notebook (replacing the model with your own), I think it should work.</p>",
      "rawMarkdown": "I think you can train the model on your own machine, then upload the model as a dataset on kaggle, and then incorporate that for you submission notebook. I think the main constraint is submitting a model instead of predictions, so if you can sort of follow the sample notebook (replacing the model with your own), I think it should work.",
      "votes": null
    },
    {
      "id": "961310",
      "postDate": "08/07/2020 05:05:18",
      "content": "<p>Yes you can</p>",
      "rawMarkdown": "Yes you can",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 926879,
      "author_name": "yushg123",
      "author_url": "",
      "post_date": "07/13/2020 04:17:07",
      "content": "<p>I think you can train the model on your own machine, then upload the model as a dataset on kaggle, and then incorporate that for you submission notebook. I think the main constraint is submitting a model instead of predictions, so if you can sort of follow the sample notebook (replacing the model with your own), I think it should work.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 961310,
      "author_name": "vilettofernando",
      "author_url": "",
      "post_date": "08/07/2020 05:05:18",
      "content": "<p>Yes you can</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "926801": "This dataset is fairly large and this puts a large constraint on the time.  Even pretrained models can't really be trained for more that 7-8 epochs.  Are we allowed to do training on our own machines?",
    "926879": "I think you can train the model on your own machine, then upload the model as a dataset on kaggle, and then incorporate that for you submission notebook. I think the main constraint is submitting a model instead of predictions, so if you can sort of follow the sample notebook (replacing the model with your own), I think it should work.",
    "961310": "Yes you can"
  },
  "source": "meta"
}