{
  "id": 214290,
  "title": "Whether to turn GPU on in my notebook when submitting if I save and run with GPU on?",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/214290",
  "author_name": "",
  "post_date": "2021-01-26T03:42:37.430393500Z",
  "votes": null,
  "comment_count": 6,
  "views": 0,
  "content": "<p>After submitting a GPU notebook, the notebook is running. I am confused whether I can turn GPU off in that notebook to save time.</p>",
  "messages": [
    {
      "id": "1170187",
      "postDate": "01/26/2021 03:42:37",
      "content": "<p>After submitting a GPU notebook, the notebook is running. I am confused whether I can turn GPU off in that notebook to save time.</p>",
      "rawMarkdown": "After submitting a GPU notebook, the notebook is running. I am confused whether I can turn GPU off in that notebook to save time.",
      "votes": null
    },
    {
      "id": "1170195",
      "postDate": "01/26/2021 03:54:30",
      "content": "<p>It's better to turn off the GPU in the notebook</p>",
      "rawMarkdown": "It's better to turn off the GPU in the notebook",
      "votes": null
    },
    {
      "id": "1170205",
      "postDate": "01/26/2021 04:08:26",
      "content": "<p>Are you talking about submitting or committing? If submitting, the run time does not count for your GPU quota. If committing, you can stop the instance if you do not need to edit the notebook or just turn off GPU to save GPU quota.</p>",
      "rawMarkdown": "Are you talking about submitting or committing? If submitting, the run time does not count for your GPU quota. If committing, you can stop the instance if you do not need to edit the notebook or just turn off GPU to save GPU quota.",
      "votes": null
    },
    {
      "id": "1170249",
      "postDate": "01/26/2021 05:01:35",
      "content": "<p>I always turn GPU on - it cost nothing towards my quota and I am not smart enough to always know what runs on GPU :)</p>\n<p>It does depend a bit on your prediction method, do you use a generator, etc.</p>\n<p>I run almost all of my code on local machines - with dual GPU.  I am always checking to see if both of the GPU's are running, only a single or none.  My observation is that at least one is running during all of my prediction kernels for this competition.</p>\n<p>I can get both GPU running during my prediction codes on several different methods (but not all).  </p>\n<p>On my main submission kernel on Kaggle I time out when I don't have the GPU running.</p>\n<p>That was long answer - short answer - using an old fashion stop watch and submit the same code with and without GPU.</p>",
      "rawMarkdown": "I always turn GPU on - it cost nothing towards my quota and I am not smart enough to always know what runs on GPU :)\n\nIt does depend a bit on your prediction method, do you use a generator, etc.\n\nI run almost all of my code on local machines - with dual GPU.  I am always checking to see if both of the GPU's are running, only a single or none.  My observation is that at least one is running during all of my prediction kernels for this competition.\n\nI can get both GPU running during my prediction codes on several different methods (but not all).  \n\nOn my main submission kernel on Kaggle I time out when I don't have the GPU running.\n\nThat was long answer - short answer - using an old fashion stop watch and submit the same code with and without GPU.",
      "votes": null
    },
    {
      "id": "1170745",
      "postDate": "01/26/2021 12:12:44",
      "content": "<p>Thank you for your patient comment.</p>",
      "rawMarkdown": "Thank you for your patient comment.",
      "votes": null
    },
    {
      "id": "1170747",
      "postDate": "01/26/2021 12:13:25",
      "content": "<p>Thank you for your helpful suggestion.</p>",
      "rawMarkdown": "Thank you for your helpful suggestion.",
      "votes": null
    },
    {
      "id": "1170748",
      "postDate": "01/26/2021 12:13:31",
      "content": "<p>Thank you for your helpful suggestion.</p>",
      "rawMarkdown": "Thank you for your helpful suggestion.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1170195,
      "author_name": "zzzhangzz",
      "author_url": "",
      "post_date": "01/26/2021 03:54:30",
      "content": "<p>It's better to turn off the GPU in the notebook</p>",
      "votes": null,
      "replies": [
        {
          "id": 1170748,
          "author_name": "lishiqian",
          "author_url": "",
          "post_date": "01/26/2021 12:13:31",
          "content": "<p>Thank you for your helpful suggestion.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1170205,
      "author_name": "woshifym",
      "author_url": "",
      "post_date": "01/26/2021 04:08:26",
      "content": "<p>Are you talking about submitting or committing? If submitting, the run time does not count for your GPU quota. If committing, you can stop the instance if you do not need to edit the notebook or just turn off GPU to save GPU quota.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1170747,
          "author_name": "lishiqian",
          "author_url": "",
          "post_date": "01/26/2021 12:13:25",
          "content": "<p>Thank you for your helpful suggestion.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1170249,
      "author_name": "pcjimmmy",
      "author_url": "",
      "post_date": "01/26/2021 05:01:35",
      "content": "<p>I always turn GPU on - it cost nothing towards my quota and I am not smart enough to always know what runs on GPU :)</p>\n<p>It does depend a bit on your prediction method, do you use a generator, etc.</p>\n<p>I run almost all of my code on local machines - with dual GPU.  I am always checking to see if both of the GPU's are running, only a single or none.  My observation is that at least one is running during all of my prediction kernels for this competition.</p>\n<p>I can get both GPU running during my prediction codes on several different methods (but not all).  </p>\n<p>On my main submission kernel on Kaggle I time out when I don't have the GPU running.</p>\n<p>That was long answer - short answer - using an old fashion stop watch and submit the same code with and without GPU.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1170745,
          "author_name": "lishiqian",
          "author_url": "",
          "post_date": "01/26/2021 12:12:44",
          "content": "<p>Thank you for your patient comment.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1170187": "After submitting a GPU notebook, the notebook is running. I am confused whether I can turn GPU off in that notebook to save time.",
    "1170195": "It's better to turn off the GPU in the notebook",
    "1170205": "Are you talking about submitting or committing? If submitting, the run time does not count for your GPU quota. If committing, you can stop the instance if you do not need to edit the notebook or just turn off GPU to save GPU quota.",
    "1170249": "I always turn GPU on - it cost nothing towards my quota and I am not smart enough to always know what runs on GPU :)\n\nIt does depend a bit on your prediction method, do you use a generator, etc.\n\nI run almost all of my code on local machines - with dual GPU.  I am always checking to see if both of the GPU's are running, only a single or none.  My observation is that at least one is running during all of my prediction kernels for this competition.\n\nI can get both GPU running during my prediction codes on several different methods (but not all).  \n\nOn my main submission kernel on Kaggle I time out when I don't have the GPU running.\n\nThat was long answer - short answer - using an old fashion stop watch and submit the same code with and without GPU.",
    "1170745": "Thank you for your patient comment.",
    "1170747": "Thank you for your helpful suggestion.",
    "1170748": "Thank you for your helpful suggestion."
  },
  "source": "meta"
}