{
  "id": 94095,
  "title": "Which kernel is faster, Script or Notebook?",
  "url": "/competitions/imet-2019-fgvc6/discussion/94095",
  "author_name": "Tawara",
  "post_date": "2019-06-02T01:30:47.719000",
  "votes": 1,
  "comment_count": 6,
  "views": 0,
  "content": "<p><a href=\"https://www.kaggle.com/c/imet-2019-fgvc6/data\">As the 2nd-stage has five times test-set of the 1st's</a>, we have to make prediction as fast as possible.</p>\n\n<p>I'm using notebook kernels now, but if script kernels are faster, I should rewrite my code for them.\nCan anyone share experiences?  </p>\n\n<p>Thanks.</p>",
  "messages": [
    {
      "id": 542046,
      "postDate": "2019-06-03T11:30:08.887Z",
      "content": "<p>I don't think there's a significant difference in inference time between them.</p>",
      "rawMarkdown": "I don't think there's a significant difference in inference time between them.",
      "votes": 1,
      "replies": [
        {
          "id": 542102,
          "postDate": "2019-06-03T13:12:21.683Z",
          "content": "<p>Thank you.\nI have only used script kernels a few times, and have the same impression.</p>",
          "rawMarkdown": "Thank you.\nI have only used script kernels a few times, and have the same impression."
        }
      ]
    },
    {
      "id": 541183,
      "postDate": "2019-06-02T01:30:47.720Z",
      "content": "<p><a href=\"https://www.kaggle.com/c/imet-2019-fgvc6/data\">As the 2nd-stage has five times test-set of the 1st's</a>, we have to make prediction as fast as possible.</p>\n\n<p>I'm using notebook kernels now, but if script kernels are faster, I should rewrite my code for them.\nCan anyone share experiences?  </p>\n\n<p>Thanks.</p>",
      "rawMarkdown": "[As the 2nd-stage has five times test-set of the 1st's](https://www.kaggle.com/c/imet-2019-fgvc6/data), we have to make prediction as fast as possible.\n\nI'm using notebook kernels now, but if script kernels are faster, I should rewrite my code for them.\nCan anyone share experiences?  \n\nThanks.",
      "votes": 1
    },
    {
      "id": 542235,
      "postDate": "2019-06-03T15:40:00.267Z",
      "content": "<p>My past experience shows that script is faster and it gives you more available GPU memory. However, you might be screwed if there is a small bug in your code (i.e., one model does not infer correctly). </p>",
      "rawMarkdown": "My past experience shows that script is faster and it gives you more available GPU memory. However, you might be screwed if there is a small bug in your code (i.e., one model does not infer correctly). ",
      "votes": 2,
      "replies": [
        {
          "id": 542268,
          "postDate": "2019-06-03T16:28:39.370Z",
          "content": "<p>Thank you for sharing. I'll try.</p>",
          "rawMarkdown": "Thank you for sharing. I'll try."
        }
      ]
    },
    {
      "id": 542233,
      "postDate": "2019-06-03T15:37:06.730Z",
      "rawMarkdown": "",
      "votes": 1,
      "isDeleted": true,
      "replies": [
        {
          "id": 542264,
          "postDate": "2019-06-03T16:25:50.513Z",
          "content": "<p>Oh, I don't know such information.\nDid you run this scripts for 9 hours?</p>",
          "rawMarkdown": "Oh, I don't know such information.\nDid you run this scripts for 9 hours?"
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 542046,
      "author_name": "seefun",
      "author_url": "",
      "post_date": "2019-06-03T11:30:08.887000",
      "content": "<p>I don't think there's a significant difference in inference time between them.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 542102,
          "author_name": "Tawara",
          "author_url": "",
          "post_date": "2019-06-03T13:12:21.683000",
          "content": "<p>Thank you.\nI have only used script kernels a few times, and have the same impression.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 542235,
      "author_name": "Xuan Cao",
      "author_url": "",
      "post_date": "2019-06-03T15:40:00.267000",
      "content": "<p>My past experience shows that script is faster and it gives you more available GPU memory. However, you might be screwed if there is a small bug in your code (i.e., one model does not infer correctly). </p>",
      "votes": 2,
      "replies": [
        {
          "id": 542268,
          "author_name": "Tawara",
          "author_url": "",
          "post_date": "2019-06-03T16:28:39.370000",
          "content": "<p>Thank you for sharing. I'll try.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 542233,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-06-03T15:37:06.730000",
      "content": "",
      "votes": 1,
      "replies": [
        {
          "id": 542264,
          "author_name": "Tawara",
          "author_url": "",
          "post_date": "2019-06-03T16:25:50.513000",
          "content": "<p>Oh, I don't know such information.\nDid you run this scripts for 9 hours?</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "542046": "I don't think there's a significant difference in inference time between them.",
    "541183": "[As the 2nd-stage has five times test-set of the 1st's](https://www.kaggle.com/c/imet-2019-fgvc6/data), we have to make prediction as fast as possible.\n\nI'm using notebook kernels now, but if script kernels are faster, I should rewrite my code for them.\nCan anyone share experiences?  \n\nThanks.",
    "542235": "My past experience shows that script is faster and it gives you more available GPU memory. However, you might be screwed if there is a small bug in your code (i.e., one model does not infer correctly). ",
    "542233": ""
  }
}