{
  "id": 182519,
  "title": "Train using Google Colab (No Pro) ",
  "url": "/competitions/lyft-motion-prediction-autonomous-vehicles/discussion/182519",
  "author_name": "",
  "post_date": "2020-09-13T07:43:36.133292800Z",
  "votes": 9,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I have made a notebook for those who want to train using Google Colab. Here is the notebook. <br>\n<a href=\"https://www.kaggle.com/kool777/ultimate-google-colab-training-batch-size-64\" target=\"_blank\">https://www.kaggle.com/kool777/ultimate-google-colab-training-batch-size-64</a></p>\n<p>Good Luck!🙌</p>",
  "messages": [
    {
      "id": "1008541",
      "postDate": "09/13/2020 07:43:36",
      "content": "<p>I have made a notebook for those who want to train using Google Colab. Here is the notebook. <br>\n<a href=\"https://www.kaggle.com/kool777/ultimate-google-colab-training-batch-size-64\" target=\"_blank\">https://www.kaggle.com/kool777/ultimate-google-colab-training-batch-size-64</a></p>\n<p>Good Luck!🙌</p>",
      "rawMarkdown": "I have made a notebook for those who want to train using Google Colab. Here is the notebook. \nhttps://www.kaggle.com/kool777/ultimate-google-colab-training-batch-size-64\n\nGood Luck!🙌",
      "votes": null
    },
    {
      "id": "1009036",
      "postDate": "09/13/2020 15:59:56",
      "content": "<p>Great work! thanks. Maybe with tpu enabled and using multiprocessing spawning on 8 cores , model could be trained in just a day instead of 7.</p>",
      "rawMarkdown": "Great work! thanks. Maybe with tpu enabled and using multiprocessing spawning on 8 cores , model could be trained in just a day instead of 7.",
      "votes": null
    },
    {
      "id": "1009384",
      "postDate": "09/14/2020 00:40:45",
      "content": "<p>This is a great idea! </p>\n<p>Could you kindly point me to any resources on porting PyTorch GPU to TPU? Thanks!</p>",
      "rawMarkdown": "This is a great idea! \n\nCould you kindly point me to any resources on porting PyTorch GPU to TPU? Thanks!",
      "votes": null
    },
    {
      "id": "1009590",
      "postDate": "09/14/2020 05:49:06",
      "content": "<p>Thank you for your suggestion. I was experimenting with GPU only. I will look for TPU as well. </p>",
      "rawMarkdown": "Thank you for your suggestion. I was experimenting with GPU only. I will look for TPU as well.",
      "votes": null
    },
    {
      "id": "1010506",
      "postDate": "09/14/2020 20:09:30",
      "content": "<p>Thanks!!!!!</p>",
      "rawMarkdown": "Thanks!!!!!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1009036,
      "author_name": "sujaydkhandekar",
      "author_url": "",
      "post_date": "09/13/2020 15:59:56",
      "content": "<p>Great work! thanks. Maybe with tpu enabled and using multiprocessing spawning on 8 cores , model could be trained in just a day instead of 7.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1009384,
          "author_name": "ltdzung",
          "author_url": "",
          "post_date": "09/14/2020 00:40:45",
          "content": "<p>This is a great idea! </p>\n<p>Could you kindly point me to any resources on porting PyTorch GPU to TPU? Thanks!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1009590,
          "author_name": "kool777",
          "author_url": "",
          "post_date": "09/14/2020 05:49:06",
          "content": "<p>Thank you for your suggestion. I was experimenting with GPU only. I will look for TPU as well. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1010506,
      "author_name": "xuejiaotang",
      "author_url": "",
      "post_date": "09/14/2020 20:09:30",
      "content": "<p>Thanks!!!!!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1008541": "I have made a notebook for those who want to train using Google Colab. Here is the notebook. \nhttps://www.kaggle.com/kool777/ultimate-google-colab-training-batch-size-64\n\nGood Luck!🙌",
    "1009036": "Great work! thanks. Maybe with tpu enabled and using multiprocessing spawning on 8 cores , model could be trained in just a day instead of 7.",
    "1009384": "This is a great idea! \n\nCould you kindly point me to any resources on porting PyTorch GPU to TPU? Thanks!",
    "1009590": "Thank you for your suggestion. I was experimenting with GPU only. I will look for TPU as well.",
    "1010506": "Thanks!!!!!"
  },
  "source": "meta"
}