{
  "id": 54633,
  "title": "Anybody using laptop for training?",
  "url": "/competitions/talkingdata-adtracking-fraud-detection/discussion/54633",
  "author_name": "",
  "post_date": "2018-04-16T06:23:48.458658100Z",
  "votes": 3,
  "comment_count": 8,
  "views": 0,
  "content": "<p>Training Dataset is a huge 1GB+ file. Is there anyone using laptop or local machine for training? </p>",
  "messages": [
    {
      "id": "314706",
      "postDate": "04/16/2018 06:23:48",
      "content": "<p>Training Dataset is a huge 1GB+ file. Is there anyone using laptop or local machine for training? </p>",
      "rawMarkdown": "Training Dataset is a huge 1GB+ file. Is there anyone using laptop or local machine for training?",
      "votes": null
    },
    {
      "id": "314707",
      "postDate": "04/16/2018 06:37:54",
      "content": "<p>Yes - 16gb MBP can handle an lgbm model on 75 mln rows.</p>",
      "rawMarkdown": "Yes - 16gb MBP can handle an lgbm model on 75 mln rows.",
      "votes": null
    },
    {
      "id": "314712",
      "postDate": "04/16/2018 06:45:46",
      "content": "<p>Time to train the model? </p>",
      "rawMarkdown": "Time to train the model?",
      "votes": null
    },
    {
      "id": "314717",
      "postDate": "04/16/2018 06:49:29",
      "content": "<p>Few hours on 8 threads ...</p>",
      "rawMarkdown": "Few hours on 8 threads ...",
      "votes": null
    },
    {
      "id": "314736",
      "postDate": "04/16/2018 07:22:05",
      "content": "<p>Thanks for information. </p>",
      "rawMarkdown": "Thanks for information.",
      "votes": null
    },
    {
      "id": "315949",
      "postDate": "04/17/2018 21:13:53",
      "content": "<p>Try R with data.table and process data in chunks. </p>\n\n<p><a href=\"https://www.kaggle.com/pranav84/lgb-entire-dataset-in-2-hrs-lb-0-9718?scriptVersionId=3218373\"> <strong>Benchmark</strong> </a> : 14 features, 4 threads, 16 GB RAM -&gt; in less than 2 hours including modelling</p>",
      "rawMarkdown": "Try R with data.table and process data in chunks. \n\n[ **Benchmark** ][1] : 14 features, 4 threads, 16 GB RAM -&gt; in less than 2 hours including modelling\n\n\n  [1]: https://www.kaggle.com/pranav84/lgb-entire-dataset-in-2-hrs-lb-0-9718?scriptVersionId=3218373",
      "votes": null
    },
    {
      "id": "315995",
      "postDate": "04/18/2018 00:44:16",
      "content": "<p>GTX1080, 32gb work station here. I ran through the data through a stateful lstm model and it took approximately 1:30:00 hours per epoch. Processing non-sequentially is about 00:10:00 per epoch. </p>",
      "rawMarkdown": "GTX1080, 32gb work station here. I ran through the data through a stateful lstm model and it took approximately 1:30:00 hours per epoch. Processing non-sequentially is about 00:10:00 per epoch.",
      "votes": null
    },
    {
      "id": "316013",
      "postDate": "04/18/2018 02:23:10",
      "content": "<p>I am using laptop with 12GB RAM and  i5-5200U , it takes 5+ hours :(</p>",
      "rawMarkdown": "I am using laptop with 12GB RAM and  i5-5200U , it takes 5+ hours :(",
      "votes": null
    },
    {
      "id": "316044",
      "postDate": "04/18/2018 04:31:15",
      "content": "<p>I’m using a 16GB RAM laptop.</p>",
      "rawMarkdown": "I’m using a 16GB RAM laptop.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 314707,
      "author_name": "konradb",
      "author_url": "",
      "post_date": "04/16/2018 06:37:54",
      "content": "<p>Yes - 16gb MBP can handle an lgbm model on 75 mln rows.</p>",
      "votes": null,
      "replies": [
        {
          "id": 314712,
          "author_name": "evan26",
          "author_url": "",
          "post_date": "04/16/2018 06:45:46",
          "content": "<p>Time to train the model? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 314717,
          "author_name": "konradb",
          "author_url": "",
          "post_date": "04/16/2018 06:49:29",
          "content": "<p>Few hours on 8 threads ...</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 314736,
          "author_name": "evan26",
          "author_url": "",
          "post_date": "04/16/2018 07:22:05",
          "content": "<p>Thanks for information. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 315949,
      "author_name": "pranav84",
      "author_url": "",
      "post_date": "04/17/2018 21:13:53",
      "content": "<p>Try R with data.table and process data in chunks. </p>\n\n<p><a href=\"https://www.kaggle.com/pranav84/lgb-entire-dataset-in-2-hrs-lb-0-9718?scriptVersionId=3218373\"> <strong>Benchmark</strong> </a> : 14 features, 4 threads, 16 GB RAM -&gt; in less than 2 hours including modelling</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 315995,
      "author_name": "vannak",
      "author_url": "",
      "post_date": "04/18/2018 00:44:16",
      "content": "<p>GTX1080, 32gb work station here. I ran through the data through a stateful lstm model and it took approximately 1:30:00 hours per epoch. Processing non-sequentially is about 00:10:00 per epoch. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 316013,
      "author_name": "dean1977",
      "author_url": "",
      "post_date": "04/18/2018 02:23:10",
      "content": "<p>I am using laptop with 12GB RAM and  i5-5200U , it takes 5+ hours :(</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 316044,
      "author_name": "tdpanalysis",
      "author_url": "",
      "post_date": "04/18/2018 04:31:15",
      "content": "<p>I’m using a 16GB RAM laptop.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "314706": "Training Dataset is a huge 1GB+ file. Is there anyone using laptop or local machine for training?",
    "314707": "Yes - 16gb MBP can handle an lgbm model on 75 mln rows.",
    "314712": "Time to train the model?",
    "314717": "Few hours on 8 threads ...",
    "314736": "Thanks for information.",
    "315949": "Try R with data.table and process data in chunks. \n\n[ **Benchmark** ][1] : 14 features, 4 threads, 16 GB RAM -&gt; in less than 2 hours including modelling\n\n\n  [1]: https://www.kaggle.com/pranav84/lgb-entire-dataset-in-2-hrs-lb-0-9718?scriptVersionId=3218373",
    "315995": "GTX1080, 32gb work station here. I ran through the data through a stateful lstm model and it took approximately 1:30:00 hours per epoch. Processing non-sequentially is about 00:10:00 per epoch.",
    "316013": "I am using laptop with 12GB RAM and  i5-5200U , it takes 5+ hours :(",
    "316044": "I’m using a 16GB RAM laptop."
  },
  "source": "meta"
}