{
  "id": 128178,
  "title": "How to submit a notebook with GPU?",
  "url": "/competitions/bengaliai-cv19/discussion/128178",
  "author_name": "",
  "post_date": "2020-01-29T11:53:21.981231800Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I have a notebook with a model, which needs 40 minutes to train on GPU (and several days to train on CPU, then)\nWhen I press \"Submit to competition\", it's sending notebook, which runs on CPU and doesn't fit into time.\nHow to submit it on GPU?</p>",
  "messages": [
    {
      "id": "732025",
      "postDate": "01/29/2020 11:53:21",
      "content": "<p>I have a notebook with a model, which needs 40 minutes to train on GPU (and several days to train on CPU, then)\nWhen I press \"Submit to competition\", it's sending notebook, which runs on CPU and doesn't fit into time.\nHow to submit it on GPU?</p>",
      "rawMarkdown": "I have a notebook with a model, which needs 40 minutes to train on GPU (and several days to train on CPU, then)\nWhen I press \"Submit to competition\", it's sending notebook, which runs on CPU and doesn't fit into time.\nHow to submit it on GPU?",
      "votes": null
    },
    {
      "id": "732044",
      "postDate": "01/29/2020 12:18:45",
      "content": "<p>I am not sure but maybe you need to turn on gpu from settings at your notebook.</p>",
      "rawMarkdown": "I am not sure but maybe you need to turn on gpu from settings at your notebook.",
      "votes": null
    },
    {
      "id": "732062",
      "postDate": "01/29/2020 12:54:31",
      "content": "<p>For this prepare two kernels, do not train while submissions.</p>\n\n<p>I created a kernel and saved it's weights and uploaded these to kernel which is used for submission.</p>",
      "rawMarkdown": "For this prepare two kernels, do not train while submissions.\n\nI created a kernel and saved it's weights and uploaded these to kernel which is used for submission.",
      "votes": null
    },
    {
      "id": "732063",
      "postDate": "01/29/2020 12:56:49",
      "content": "<p>Have a look at it this is my first kernel I am saving weights here.</p>\n\n<p>And used this weights in second kernel and not training again on submission kernel</p>\n\n<p><a href=\"https://www.kaggle.com/amit9484/keras-cnn-saving-model-lb-score-89\">https://www.kaggle.com/amit9484/keras-cnn-saving-model-lb-score-89</a></p>",
      "rawMarkdown": "Have a look at it this is my first kernel I am saving weights here.\n\nAnd used this weights in second kernel and not training again on submission kernel\n\nhttps://www.kaggle.com/amit9484/keras-cnn-saving-model-lb-score-89",
      "votes": null
    },
    {
      "id": "733519",
      "postDate": "01/31/2020 08:49:26",
      "content": "<p>I think its much better to save the Model and run the inference on saved Model, then it becomes a problem of Memory management only, I have done that way, its working good</p>",
      "rawMarkdown": "I think its much better to save the Model and run the inference on saved Model, then it becomes a problem of Memory management only, I have done that way, its working good",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 732044,
      "author_name": "erkanhatipoglu",
      "author_url": "",
      "post_date": "01/29/2020 12:18:45",
      "content": "<p>I am not sure but maybe you need to turn on gpu from settings at your notebook.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 732062,
      "author_name": "amit9484",
      "author_url": "",
      "post_date": "01/29/2020 12:54:31",
      "content": "<p>For this prepare two kernels, do not train while submissions.</p>\n\n<p>I created a kernel and saved it's weights and uploaded these to kernel which is used for submission.</p>",
      "votes": null,
      "replies": [
        {
          "id": 732063,
          "author_name": "amit9484",
          "author_url": "",
          "post_date": "01/29/2020 12:56:49",
          "content": "<p>Have a look at it this is my first kernel I am saving weights here.</p>\n\n<p>And used this weights in second kernel and not training again on submission kernel</p>\n\n<p><a href=\"https://www.kaggle.com/amit9484/keras-cnn-saving-model-lb-score-89\">https://www.kaggle.com/amit9484/keras-cnn-saving-model-lb-score-89</a></p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 733519,
      "author_name": "anirbank",
      "author_url": "",
      "post_date": "01/31/2020 08:49:26",
      "content": "<p>I think its much better to save the Model and run the inference on saved Model, then it becomes a problem of Memory management only, I have done that way, its working good</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "732025": "I have a notebook with a model, which needs 40 minutes to train on GPU (and several days to train on CPU, then)\nWhen I press \"Submit to competition\", it's sending notebook, which runs on CPU and doesn't fit into time.\nHow to submit it on GPU?",
    "732044": "I am not sure but maybe you need to turn on gpu from settings at your notebook.",
    "732062": "For this prepare two kernels, do not train while submissions.\n\nI created a kernel and saved it's weights and uploaded these to kernel which is used for submission.",
    "732063": "Have a look at it this is my first kernel I am saving weights here.\n\nAnd used this weights in second kernel and not training again on submission kernel\n\nhttps://www.kaggle.com/amit9484/keras-cnn-saving-model-lb-score-89",
    "733519": "I think its much better to save the Model and run the inference on saved Model, then it becomes a problem of Memory management only, I have done that way, its working good"
  },
  "source": "meta"
}