{
  "id": 231794,
  "title": "\"Your Notebook cannot use internet access in this competition\" and other questions",
  "url": "/competitions/plant-pathology-2021-fgvc8/discussion/231794",
  "author_name": "",
  "post_date": "2021-04-10T11:20:19.501298600Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I use the Mobilenet model with ImageNet pre-trained parameters. In the beginning, the Notebook will download these parameters, but I cannot submit the code due to \"Your Notebook cannot use internet access in this competition\". If I disable the Internet, the Notebook cannot download parameters, hence my code cannot execute. How can I solve this problem?</p>\n<p>Where should I save the submission file? In the \"/kaggle/output\" directory, right?</p>\n<p>If it possible, please guide me step-by-step on how I can submit my Notebook. I just used simple Mobilenet with pre-trained model.</p>\n<p>Mayny thanks!</p>",
  "messages": [
    {
      "id": "1269263",
      "postDate": "04/10/2021 11:20:19",
      "content": "<p>I use the Mobilenet model with ImageNet pre-trained parameters. In the beginning, the Notebook will download these parameters, but I cannot submit the code due to \"Your Notebook cannot use internet access in this competition\". If I disable the Internet, the Notebook cannot download parameters, hence my code cannot execute. How can I solve this problem?</p>\n<p>Where should I save the submission file? In the \"/kaggle/output\" directory, right?</p>\n<p>If it possible, please guide me step-by-step on how I can submit my Notebook. I just used simple Mobilenet with pre-trained model.</p>\n<p>Mayny thanks!</p>",
      "rawMarkdown": "I use the Mobilenet model with ImageNet pre-trained parameters. In the beginning, the Notebook will download these parameters, but I cannot submit the code due to \"Your Notebook cannot use internet access in this competition\". If I disable the Internet, the Notebook cannot download parameters, hence my code cannot execute. How can I solve this problem?\n\nWhere should I save the submission file? In the \"/kaggle/output\" directory, right?\n\nIf it possible, please guide me step-by-step on how I can submit my Notebook. I just used simple Mobilenet with pre-trained model.\n\nMayny thanks!",
      "votes": null
    },
    {
      "id": "1269426",
      "postDate": "04/10/2021 14:21:48",
      "content": "<p>You need to split your notebook to two notebooks: training and inference. You can use internet in the first one - use it to train model and save your weights. Create a dataset with that weights, and then use it in the second notebook with disabled Internet which you can submit. If you're using Tensorflow, use <code>weights=None</code> paramater to disable weights download, and use <code>model.load_weights(...)</code> to load weights to the model. </p>",
      "rawMarkdown": "You need to split your notebook to two notebooks: training and inference. You can use internet in the first one - use it to train model and save your weights. Create a dataset with that weights, and then use it in the second notebook with disabled Internet which you can submit. If you're using Tensorflow, use `weights=None` paramater to disable weights download, and use `model.load_weights(...)` to load weights to the model.",
      "votes": null
    },
    {
      "id": "1269508",
      "postDate": "04/10/2021 15:45:59",
      "content": "<p>So basically we have 2 steps:<br>\nStep 1: Enable Internet. Download and save the pre-train model.<br>\nStep 2: Disable Internet, set weights=None to disable weights download, and load the downloaded pre-train model.</p>\n<p>Thank you.</p>",
      "rawMarkdown": "So basically we have 2 steps:\nStep 1: Enable Internet. Download and save the pre-train model.\nStep 2: Disable Internet, set weights=None to disable weights download, and load the downloaded pre-train model.\n\nThank you.",
      "votes": null
    },
    {
      "id": "1274237",
      "postDate": "04/15/2021 06:08:54",
      "content": "<p>In prediction notebook, if you have a problem \"unknown layer * \". you can download model layer <a href=\"https://www.kaggle.com/xhlulu/keras-efficientnet-noisy-students\" target=\"_blank\">https://www.kaggle.com/xhlulu/keras-efficientnet-noisy-students</a>. </p>",
      "rawMarkdown": "In prediction notebook, if you have a problem \"unknown layer * \". you can download model layer https://www.kaggle.com/xhlulu/keras-efficientnet-noisy-students.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1269426,
      "author_name": "atamazian",
      "author_url": "",
      "post_date": "04/10/2021 14:21:48",
      "content": "<p>You need to split your notebook to two notebooks: training and inference. You can use internet in the first one - use it to train model and save your weights. Create a dataset with that weights, and then use it in the second notebook with disabled Internet which you can submit. If you're using Tensorflow, use <code>weights=None</code> paramater to disable weights download, and use <code>model.load_weights(...)</code> to load weights to the model. </p>",
      "votes": null,
      "replies": [
        {
          "id": 1269508,
          "author_name": "hoangtrong2305",
          "author_url": "",
          "post_date": "04/10/2021 15:45:59",
          "content": "<p>So basically we have 2 steps:<br>\nStep 1: Enable Internet. Download and save the pre-train model.<br>\nStep 2: Disable Internet, set weights=None to disable weights download, and load the downloaded pre-train model.</p>\n<p>Thank you.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1274237,
      "author_name": "hellolion",
      "author_url": "",
      "post_date": "04/15/2021 06:08:54",
      "content": "<p>In prediction notebook, if you have a problem \"unknown layer * \". you can download model layer <a href=\"https://www.kaggle.com/xhlulu/keras-efficientnet-noisy-students\" target=\"_blank\">https://www.kaggle.com/xhlulu/keras-efficientnet-noisy-students</a>. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1269263": "I use the Mobilenet model with ImageNet pre-trained parameters. In the beginning, the Notebook will download these parameters, but I cannot submit the code due to \"Your Notebook cannot use internet access in this competition\". If I disable the Internet, the Notebook cannot download parameters, hence my code cannot execute. How can I solve this problem?\n\nWhere should I save the submission file? In the \"/kaggle/output\" directory, right?\n\nIf it possible, please guide me step-by-step on how I can submit my Notebook. I just used simple Mobilenet with pre-trained model.\n\nMayny thanks!",
    "1269426": "You need to split your notebook to two notebooks: training and inference. You can use internet in the first one - use it to train model and save your weights. Create a dataset with that weights, and then use it in the second notebook with disabled Internet which you can submit. If you're using Tensorflow, use `weights=None` paramater to disable weights download, and use `model.load_weights(...)` to load weights to the model.",
    "1269508": "So basically we have 2 steps:\nStep 1: Enable Internet. Download and save the pre-train model.\nStep 2: Disable Internet, set weights=None to disable weights download, and load the downloaded pre-train model.\n\nThank you.",
    "1274237": "In prediction notebook, if you have a problem \"unknown layer * \". you can download model layer https://www.kaggle.com/xhlulu/keras-efficientnet-noisy-students."
  },
  "source": "meta"
}