{
  "id": 206505,
  "title": "Is it possible to submit a kernel with GPU",
  "url": "/competitions/riiid-test-answer-prediction/discussion/206505",
  "author_name": "",
  "post_date": "2020-12-25T01:14:26.143240600Z",
  "votes": 1,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Inference using Transformer takes long time using CPU, but way less on a GPU. Is it possible to submit a kernel with GPU turned on? My kernels are timing out because of this!</p>",
  "messages": [
    {
      "id": "1125698",
      "postDate": "12/25/2020 01:14:26",
      "content": "<p>Inference using Transformer takes long time using CPU, but way less on a GPU. Is it possible to submit a kernel with GPU turned on? My kernels are timing out because of this!</p>",
      "rawMarkdown": "Inference using Transformer takes long time using CPU, but way less on a GPU. Is it possible to submit a kernel with GPU turned on? My kernels are timing out because of this!",
      "votes": null
    },
    {
      "id": "1125824",
      "postDate": "12/25/2020 05:29:23",
      "content": "<p>Yes it is possible. While saving and committing the notebook go into advanced settings and select the run this notebook with accelerator option.</p>",
      "rawMarkdown": "Yes it is possible. While saving and committing the notebook go into advanced settings and select the run this notebook with accelerator option.",
      "votes": null
    },
    {
      "id": "1125918",
      "postDate": "12/25/2020 07:12:15",
      "content": "<p>Yes it is possible. Here are a few notes about the Kaggle virtual machine with GPU turned on and off</p>\n<ul>\n<li>GPU <strong>on</strong> : CPU has 2 cores, CPU has 13GB memory, GPU has 16 GB memory</li>\n<li>GPU <strong>off</strong>: CPU has 4 cores, CPU has 16GB memory, GPU off</li>\n</ul>",
      "rawMarkdown": "Yes it is possible. Here are a few notes about the Kaggle virtual machine with GPU turned on and off\n* GPU **on** : CPU has 2 cores, CPU has 13GB memory, GPU has 16 GB memory\n* GPU **off**: CPU has 4 cores, CPU has 16GB memory, GPU off",
      "votes": null
    },
    {
      "id": "1126175",
      "postDate": "12/25/2020 12:06:21",
      "content": "<p>I thought that turns on GPU only for committing ?</p>",
      "rawMarkdown": "I thought that turns on GPU only for committing ?",
      "votes": null
    },
    {
      "id": "1126306",
      "postDate": "12/25/2020 14:06:50",
      "content": "<p>But how to turned it on ? When I commit my notebook -&gt; advanced settings -&gt; Run with GPU, but when I submit my notebook it behaves as if it is using CPU and times out. Is this normal ?</p>",
      "rawMarkdown": "But how to turned it on ? When I commit my notebook -> advanced settings -> Run with GPU, but when I submit my notebook it behaves as if it is using CPU and times out. Is this normal ?",
      "votes": null
    },
    {
      "id": "1126311",
      "postDate": "12/25/2020 14:10:24",
      "content": "<p>It shouldn't time out with GPU turned on. Select the option run with GPU for all sessions and also select save the output. If this is not even working, then try to fine-tune your code or contact the Kaggle support team</p>",
      "rawMarkdown": "It shouldn't time out with GPU turned on. Select the option run with GPU for all sessions and also select save the output. If this is not even working, then try to fine-tune your code or contact the Kaggle support team",
      "votes": null
    },
    {
      "id": "1126316",
      "postDate": "12/25/2020 14:14:01",
      "content": "<p><a href=\"https://www.kaggle.com/atharvaingle\" target=\"_blank\">@atharvaingle</a> Thanks for the follow up, it is not working indeed. This is confusing me so much, I am sure my code work. I contacted kaggle and I'm still waiting for a response. <br>\nSuch a bad timing for such an error.</p>",
      "rawMarkdown": "atharvaingle Thanks for the follow up, it is not working indeed. This is confusing me so much, I am sure my code work. I contacted kaggle and I'm still waiting for a response. \nSuch a bad timing for such an error.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1125824,
      "author_name": "atharvaingle",
      "author_url": "",
      "post_date": "12/25/2020 05:29:23",
      "content": "<p>Yes it is possible. While saving and committing the notebook go into advanced settings and select the run this notebook with accelerator option.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1126175,
          "author_name": "abdessalemboukil",
          "author_url": "",
          "post_date": "12/25/2020 12:06:21",
          "content": "<p>I thought that turns on GPU only for committing ?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1125918,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "12/25/2020 07:12:15",
      "content": "<p>Yes it is possible. Here are a few notes about the Kaggle virtual machine with GPU turned on and off</p>\n<ul>\n<li>GPU <strong>on</strong> : CPU has 2 cores, CPU has 13GB memory, GPU has 16 GB memory</li>\n<li>GPU <strong>off</strong>: CPU has 4 cores, CPU has 16GB memory, GPU off</li>\n</ul>",
      "votes": null,
      "replies": [
        {
          "id": 1126306,
          "author_name": "abdessalemboukil",
          "author_url": "",
          "post_date": "12/25/2020 14:06:50",
          "content": "<p>But how to turned it on ? When I commit my notebook -&gt; advanced settings -&gt; Run with GPU, but when I submit my notebook it behaves as if it is using CPU and times out. Is this normal ?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1126311,
          "author_name": "atharvaingle",
          "author_url": "",
          "post_date": "12/25/2020 14:10:24",
          "content": "<p>It shouldn't time out with GPU turned on. Select the option run with GPU for all sessions and also select save the output. If this is not even working, then try to fine-tune your code or contact the Kaggle support team</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1126316,
          "author_name": "abdessalemboukil",
          "author_url": "",
          "post_date": "12/25/2020 14:14:01",
          "content": "<p><a href=\"https://www.kaggle.com/atharvaingle\" target=\"_blank\">@atharvaingle</a> Thanks for the follow up, it is not working indeed. This is confusing me so much, I am sure my code work. I contacted kaggle and I'm still waiting for a response. <br>\nSuch a bad timing for such an error.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1125698": "Inference using Transformer takes long time using CPU, but way less on a GPU. Is it possible to submit a kernel with GPU turned on? My kernels are timing out because of this!",
    "1125824": "Yes it is possible. While saving and committing the notebook go into advanced settings and select the run this notebook with accelerator option.",
    "1125918": "Yes it is possible. Here are a few notes about the Kaggle virtual machine with GPU turned on and off\n* GPU **on** : CPU has 2 cores, CPU has 13GB memory, GPU has 16 GB memory\n* GPU **off**: CPU has 4 cores, CPU has 16GB memory, GPU off",
    "1126175": "I thought that turns on GPU only for committing ?",
    "1126306": "But how to turned it on ? When I commit my notebook -> advanced settings -> Run with GPU, but when I submit my notebook it behaves as if it is using CPU and times out. Is this normal ?",
    "1126311": "It shouldn't time out with GPU turned on. Select the option run with GPU for all sessions and also select save the output. If this is not even working, then try to fine-tune your code or contact the Kaggle support team",
    "1126316": "atharvaingle Thanks for the follow up, it is not working indeed. This is confusing me so much, I am sure my code work. I contacted kaggle and I'm still waiting for a response. \nSuch a bad timing for such an error."
  },
  "source": "meta"
}