{
  "id": 125359,
  "title": "How to train for long hours in GPU in Kaggle kernels?",
  "url": "/competitions/bengaliai-cv19/discussion/125359",
  "author_name": "",
  "post_date": "2020-01-10T05:53:53.150515400Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>It's mentioned in the rules that:\nGPU Notebook &lt;= 2 hours run-time for submission</p>",
  "messages": [
    {
      "id": "715094",
      "postDate": "01/10/2020 05:53:53",
      "content": "<p>It's mentioned in the rules that:\nGPU Notebook &lt;= 2 hours run-time for submission</p>",
      "rawMarkdown": "It's mentioned in the rules that:\nGPU Notebook &lt;= 2 hours run-time for submission",
      "votes": null
    },
    {
      "id": "715101",
      "postDate": "01/10/2020 06:05:26",
      "content": "<blockquote>\n  <p>train in one kernel(for two hours) -&gt; save model -&gt; load model in another kernel -&gt; resume training</p>\n</blockquote>",
      "rawMarkdown": "&gt; train in one kernel(for two hours) -&gt; save model -&gt; load model in another kernel -&gt; resume training",
      "votes": null
    },
    {
      "id": "717900",
      "postDate": "01/13/2020 18:53:29",
      "content": "<p>Thanks <a href=\"/bibek777\">@bibek777</a> </p>",
      "rawMarkdown": "Thanks @bibek777",
      "votes": null
    },
    {
      "id": "718735",
      "postDate": "01/14/2020 17:42:29",
      "content": "<p>Earlier when kaggle didn't have GPU quota, I used kernels+bitbucket (github won't allow uploading large models) even for my personal projects\n- git clone\n- run training till kernel times out\n- git commit\n- git clone again\n- run training (it should continue) until time out\n- git commit\n....</p>",
      "rawMarkdown": "Earlier when kaggle didn't have GPU quota, I used kernels+bitbucket (github won't allow uploading large models) even for my personal projects\n- git clone\n- run training till kernel times out\n- git commit\n- git clone again\n- run training (it should continue) until time out\n- git commit\n....",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 715101,
      "author_name": "bibek777",
      "author_url": "",
      "post_date": "01/10/2020 06:05:26",
      "content": "<blockquote>\n  <p>train in one kernel(for two hours) -&gt; save model -&gt; load model in another kernel -&gt; resume training</p>\n</blockquote>",
      "votes": null,
      "replies": []
    },
    {
      "id": 717900,
      "author_name": "vishnurapps",
      "author_url": "",
      "post_date": "01/13/2020 18:53:29",
      "content": "<p>Thanks <a href=\"/bibek777\">@bibek777</a> </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 718735,
      "author_name": "timetraveller98",
      "author_url": "",
      "post_date": "01/14/2020 17:42:29",
      "content": "<p>Earlier when kaggle didn't have GPU quota, I used kernels+bitbucket (github won't allow uploading large models) even for my personal projects\n- git clone\n- run training till kernel times out\n- git commit\n- git clone again\n- run training (it should continue) until time out\n- git commit\n....</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "715094": "It's mentioned in the rules that:\nGPU Notebook &lt;= 2 hours run-time for submission",
    "715101": "&gt; train in one kernel(for two hours) -&gt; save model -&gt; load model in another kernel -&gt; resume training",
    "717900": "Thanks @bibek777",
    "718735": "Earlier when kaggle didn't have GPU quota, I used kernels+bitbucket (github won't allow uploading large models) even for my personal projects\n- git clone\n- run training till kernel times out\n- git commit\n- git clone again\n- run training (it should continue) until time out\n- git commit\n...."
  },
  "source": "meta"
}