{
  "id": 116907,
  "title": "Trouble with training time",
  "url": "/competitions/Kannada-MNIST/discussion/116907",
  "author_name": "",
  "post_date": "2019-11-12T07:52:12.522132400Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi all, I've adapted code from my original MNIST kernel, but I'm having some trouble with running times. The training seems super slow!</p>\n\n<p>My kernel is the following one:\n<a href=\"https://www.kaggle.com/amneves/mnist-with-tensorflow-keras-cnn\">Kernel here</a></p>\n\n<p>Seems odd to take this long with only 60k training samples, and my network is not that big...\nAny advice?</p>\n\n<p>I added early stopping to see if it improves, but would like to speed up per epoch!</p>\n\n<p>Thank you in advance, feel free to criticize it's how we learn!</p>",
  "messages": [
    {
      "id": "671061",
      "postDate": "11/12/2019 07:52:12",
      "content": "<p>Hi all, I've adapted code from my original MNIST kernel, but I'm having some trouble with running times. The training seems super slow!</p>\n\n<p>My kernel is the following one:\n<a href=\"https://www.kaggle.com/amneves/mnist-with-tensorflow-keras-cnn\">Kernel here</a></p>\n\n<p>Seems odd to take this long with only 60k training samples, and my network is not that big...\nAny advice?</p>\n\n<p>I added early stopping to see if it improves, but would like to speed up per epoch!</p>\n\n<p>Thank you in advance, feel free to criticize it's how we learn!</p>",
      "rawMarkdown": "Hi all, I've adapted code from my original MNIST kernel, but I'm having some trouble with running times. The training seems super slow!\n\nMy kernel is the following one:\n[Kernel here](https://www.kaggle.com/amneves/mnist-with-tensorflow-keras-cnn)\n\nSeems odd to take this long with only 60k training samples, and my network is not that big...\nAny advice?\n\nI added early stopping to see if it improves, but would like to speed up per epoch!\n\nThank you in advance, feel free to criticize it's how we learn!",
      "votes": null
    },
    {
      "id": "671408",
      "postDate": "11/12/2019 16:47:55",
      "content": "<p>Have you enabled the gpu? The run time shouldn't be that much different to the original MNIST kernel. The training set in that competition is 42k records I believe and here it is 60k so the difference in speed should be in that kind of ratio other things being equal.</p>\n\n<p>If you are editing your kernel the gpu status should be visible to the bottom right of your screen. If it isn't then you need to click the arrow pointing left in the top right of your screen, this should make it visible in the bottom right.</p>",
      "rawMarkdown": "Have you enabled the gpu? The run time shouldn't be that much different to the original MNIST kernel. The training set in that competition is 42k records I believe and here it is 60k so the difference in speed should be in that kind of ratio other things being equal.\n\nIf you are editing your kernel the gpu status should be visible to the bottom right of your screen. If it isn't then you need to click the arrow pointing left in the top right of your screen, this should make it visible in the bottom right.",
      "votes": null
    },
    {
      "id": "671596",
      "postDate": "11/12/2019 23:26:50",
      "content": "<p>I think I had it enabled, I blew my GPU quota on that kernel.</p>\n\n<p>Ill wait for it to come back and try again. Thanks</p>",
      "rawMarkdown": "I think I had it enabled, I blew my GPU quota on that kernel.\n\nIll wait for it to come back and try again. Thanks",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 671408,
      "author_name": "datahobbit",
      "author_url": "",
      "post_date": "11/12/2019 16:47:55",
      "content": "<p>Have you enabled the gpu? The run time shouldn't be that much different to the original MNIST kernel. The training set in that competition is 42k records I believe and here it is 60k so the difference in speed should be in that kind of ratio other things being equal.</p>\n\n<p>If you are editing your kernel the gpu status should be visible to the bottom right of your screen. If it isn't then you need to click the arrow pointing left in the top right of your screen, this should make it visible in the bottom right.</p>",
      "votes": null,
      "replies": [
        {
          "id": 671596,
          "author_name": "amneves",
          "author_url": "",
          "post_date": "11/12/2019 23:26:50",
          "content": "<p>I think I had it enabled, I blew my GPU quota on that kernel.</p>\n\n<p>Ill wait for it to come back and try again. Thanks</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "671061": "Hi all, I've adapted code from my original MNIST kernel, but I'm having some trouble with running times. The training seems super slow!\n\nMy kernel is the following one:\n[Kernel here](https://www.kaggle.com/amneves/mnist-with-tensorflow-keras-cnn)\n\nSeems odd to take this long with only 60k training samples, and my network is not that big...\nAny advice?\n\nI added early stopping to see if it improves, but would like to speed up per epoch!\n\nThank you in advance, feel free to criticize it's how we learn!",
    "671408": "Have you enabled the gpu? The run time shouldn't be that much different to the original MNIST kernel. The training set in that competition is 42k records I believe and here it is 60k so the difference in speed should be in that kind of ratio other things being equal.\n\nIf you are editing your kernel the gpu status should be visible to the bottom right of your screen. If it isn't then you need to click the arrow pointing left in the top right of your screen, this should make it visible in the bottom right.",
    "671596": "I think I had it enabled, I blew my GPU quota on that kernel.\n\nIll wait for it to come back and try again. Thanks"
  },
  "source": "meta"
}