{
  "id": 75115,
  "title": "GPU On vs. Off",
  "url": "/competitions/quora-insincere-questions-classification/discussion/75115",
  "author_name": "",
  "post_date": "2018-12-18T16:51:03.573388500Z",
  "votes": -1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>In the kernel console, next to CPU usage, there is a indicator for GPU on / off.    How do I turn GPU on and would I see any improvement to run speed as a result?</p>",
  "messages": [
    {
      "id": "441432",
      "postDate": "12/18/2018 16:51:03",
      "content": "<p>In the kernel console, next to CPU usage, there is a indicator for GPU on / off.    How do I turn GPU on and would I see any improvement to run speed as a result?</p>",
      "rawMarkdown": "In the kernel console, next to CPU usage, there is a indicator for GPU on / off.    How do I turn GPU on and would I see any improvement to run speed as a result?",
      "votes": null
    },
    {
      "id": "441442",
      "postDate": "12/18/2018 17:05:14",
      "content": "<p>On the same screen, below the CPU usage info towards the bottom on the screen, you can find a setting called \"GPU\" where you can switch it on or off.</p>\n\n<p>GPU is faster when using CuDNN. In Keras, there are the <code>CuDNNLSTM</code> and <code>CuDNNGRU</code> layers that you can use. They are much faster, but have some other disadvantages (e.g. no recurrent dropout, high randomness, etc.).</p>\n\n<p>Read more here: <a href=\"https://keras.io/layers/recurrent/\">https://keras.io/layers/recurrent/</a></p>",
      "rawMarkdown": "On the same screen, below the CPU usage info towards the bottom on the screen, you can find a setting called \"GPU\" where you can switch it on or off.\n\nGPU is faster when using CuDNN. In Keras, there are the `CuDNNLSTM` and `CuDNNGRU` layers that you can use. They are much faster, but have some other disadvantages (e.g. no recurrent dropout, high randomness, etc.).\n\nRead more here: https://keras.io/layers/recurrent/",
      "votes": null
    },
    {
      "id": "441548",
      "postDate": "12/18/2018 19:25:50",
      "content": "<p>Could you give the cue, please, what is the reason for no possibility to switch \"GPU\" on? \nI see just \"GPU off\" and the setting called \"GPU\" doesn't respond to my click. </p>",
      "rawMarkdown": "Could you give the cue, please, what is the reason for no possibility to switch \"GPU\" on? \nI see just \"GPU off\" and the setting called \"GPU\" doesn't respond to my click.",
      "votes": null
    },
    {
      "id": "441581",
      "postDate": "12/18/2018 20:03:27",
      "content": "<p>Here, I made a screenshot of the option. If that doesn't work for you, something else must be fishy. Consider contacting Kaggle.</p>\n\n<p><img src=\"http://www.maxschumacher.info/misc/kaggle.jpg\" alt=\"Image\"></p>",
      "rawMarkdown": "Here, I made a screenshot of the option. If that doesn't work for you, something else must be fishy. Consider contacting Kaggle.\n\n![Image](http://www.maxschumacher.info/misc/kaggle.jpg)",
      "votes": null
    },
    {
      "id": "441615",
      "postDate": "12/18/2018 20:57:50",
      "content": "<p>Thank you!</p>",
      "rawMarkdown": "Thank you!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 441442,
      "author_name": "mschumacher",
      "author_url": "",
      "post_date": "12/18/2018 17:05:14",
      "content": "<p>On the same screen, below the CPU usage info towards the bottom on the screen, you can find a setting called \"GPU\" where you can switch it on or off.</p>\n\n<p>GPU is faster when using CuDNN. In Keras, there are the <code>CuDNNLSTM</code> and <code>CuDNNGRU</code> layers that you can use. They are much faster, but have some other disadvantages (e.g. no recurrent dropout, high randomness, etc.).</p>\n\n<p>Read more here: <a href=\"https://keras.io/layers/recurrent/\">https://keras.io/layers/recurrent/</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 441548,
          "author_name": "natalyr",
          "author_url": "",
          "post_date": "12/18/2018 19:25:50",
          "content": "<p>Could you give the cue, please, what is the reason for no possibility to switch \"GPU\" on? \nI see just \"GPU off\" and the setting called \"GPU\" doesn't respond to my click. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 441581,
          "author_name": "mschumacher",
          "author_url": "",
          "post_date": "12/18/2018 20:03:27",
          "content": "<p>Here, I made a screenshot of the option. If that doesn't work for you, something else must be fishy. Consider contacting Kaggle.</p>\n\n<p><img src=\"http://www.maxschumacher.info/misc/kaggle.jpg\" alt=\"Image\"></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 441615,
          "author_name": "natalyr",
          "author_url": "",
          "post_date": "12/18/2018 20:57:50",
          "content": "<p>Thank you!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "441432": "In the kernel console, next to CPU usage, there is a indicator for GPU on / off.    How do I turn GPU on and would I see any improvement to run speed as a result?",
    "441442": "On the same screen, below the CPU usage info towards the bottom on the screen, you can find a setting called \"GPU\" where you can switch it on or off.\n\nGPU is faster when using CuDNN. In Keras, there are the `CuDNNLSTM` and `CuDNNGRU` layers that you can use. They are much faster, but have some other disadvantages (e.g. no recurrent dropout, high randomness, etc.).\n\nRead more here: https://keras.io/layers/recurrent/",
    "441548": "Could you give the cue, please, what is the reason for no possibility to switch \"GPU\" on? \nI see just \"GPU off\" and the setting called \"GPU\" doesn't respond to my click.",
    "441581": "Here, I made a screenshot of the option. If that doesn't work for you, something else must be fishy. Consider contacting Kaggle.\n\n![Image](http://www.maxschumacher.info/misc/kaggle.jpg)",
    "441615": "Thank you!"
  },
  "source": "meta"
}