{
  "id": 466683,
  "title": "Resolving external dependencies",
  "url": "/competitions/blood-vessel-segmentation/discussion/466683",
  "author_name": "",
  "post_date": "2024-01-09T15:11:24.664297800Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello, </p>\n<p>I am a bit late but I am training my first model on the competition just now. I guess this may be <br>\na silly question, and I am sorry in advance if this is the case.</p>\n<p>For training and inference I was planning to use the <code>monai</code> library. <br>\nI saw that it is possible to perform training offline, save the model's weights and then submit <br>\nanother notebook with the predictions done by loading the previously saved model.<br>\nIn addition, the submission notebooks must be \"offline\" when submitted, thus we cannot use any other external library for inference. <br>\nIf this is the case, there is no problem with the training part. However, in order to submit my inference notebook, I would need the forbidden internet access, thus I would need to rewrite my code by just using pytorch.</p>\n<p>Is this correct or am I missing something? </p>\n<p>Thank you very much for your clarification.</p>\n<p>Best regards</p>",
  "messages": [
    {
      "id": "2593944",
      "postDate": "01/09/2024 15:11:24",
      "content": "<p>Hello, </p>\n<p>I am a bit late but I am training my first model on the competition just now. I guess this may be <br>\na silly question, and I am sorry in advance if this is the case.</p>\n<p>For training and inference I was planning to use the <code>monai</code> library. <br>\nI saw that it is possible to perform training offline, save the model's weights and then submit <br>\nanother notebook with the predictions done by loading the previously saved model.<br>\nIn addition, the submission notebooks must be \"offline\" when submitted, thus we cannot use any other external library for inference. <br>\nIf this is the case, there is no problem with the training part. However, in order to submit my inference notebook, I would need the forbidden internet access, thus I would need to rewrite my code by just using pytorch.</p>\n<p>Is this correct or am I missing something? </p>\n<p>Thank you very much for your clarification.</p>\n<p>Best regards</p>",
      "rawMarkdown": "Hello, \n\nI am a bit late but I am training my first model on the competition just now. I guess this may be \na silly question, and I am sorry in advance if this is the case.\n\nFor training and inference I was planning to use the `monai` library. \nI saw that it is possible to perform training offline, save the model's weights and then submit \nanother notebook with the predictions done by loading the previously saved model.\nIn addition, the submission notebooks must be \"offline\" when submitted, thus we cannot use any other external library for inference. \nIf this is the case, there is no problem with the training part. However, in order to submit my inference notebook, I would need the forbidden internet access, thus I would need to rewrite my code by just using pytorch.\n\nIs this correct or am I missing something? \n\nThank you very much for your clarification.\n\nBest regards",
      "votes": null
    },
    {
      "id": "2594020",
      "postDate": "01/09/2024 16:05:39",
      "content": "<p>You can upload the installation of external dependencies as dataset and install them offline. I've seen the process for SMP and cc3d. Not sure about monai.</p>",
      "rawMarkdown": "You can upload the installation of external dependencies as dataset and install them offline. I've seen the process for SMP and cc3d. Not sure about monai.",
      "votes": null
    },
    {
      "id": "2595314",
      "postDate": "01/10/2024 10:58:58",
      "content": "<p>Thanks! That solved my problem! Here's the link for the tutorial I found, if anyone else was wondering the same thing <a href=\"https://www.kaggle.com/code/truthr/install-python-packages-without-internet-access/notebook\" target=\"_blank\">https://www.kaggle.com/code/truthr/install-python-packages-without-internet-access/notebook</a></p>",
      "rawMarkdown": "Thanks! That solved my problem! Here's the link for the tutorial I found, if anyone else was wondering the same thing https://www.kaggle.com/code/truthr/install-python-packages-without-internet-access/notebook",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2594020,
      "author_name": "sacuscreed",
      "author_url": "",
      "post_date": "01/09/2024 16:05:39",
      "content": "<p>You can upload the installation of external dependencies as dataset and install them offline. I've seen the process for SMP and cc3d. Not sure about monai.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2595314,
          "author_name": "fmgsf12",
          "author_url": "",
          "post_date": "01/10/2024 10:58:58",
          "content": "<p>Thanks! That solved my problem! Here's the link for the tutorial I found, if anyone else was wondering the same thing <a href=\"https://www.kaggle.com/code/truthr/install-python-packages-without-internet-access/notebook\" target=\"_blank\">https://www.kaggle.com/code/truthr/install-python-packages-without-internet-access/notebook</a></p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2593944": "Hello, \n\nI am a bit late but I am training my first model on the competition just now. I guess this may be \na silly question, and I am sorry in advance if this is the case.\n\nFor training and inference I was planning to use the `monai` library. \nI saw that it is possible to perform training offline, save the model's weights and then submit \nanother notebook with the predictions done by loading the previously saved model.\nIn addition, the submission notebooks must be \"offline\" when submitted, thus we cannot use any other external library for inference. \nIf this is the case, there is no problem with the training part. However, in order to submit my inference notebook, I would need the forbidden internet access, thus I would need to rewrite my code by just using pytorch.\n\nIs this correct or am I missing something? \n\nThank you very much for your clarification.\n\nBest regards",
    "2594020": "You can upload the installation of external dependencies as dataset and install them offline. I've seen the process for SMP and cc3d. Not sure about monai.",
    "2595314": "Thanks! That solved my problem! Here's the link for the tutorial I found, if anyone else was wondering the same thing https://www.kaggle.com/code/truthr/install-python-packages-without-internet-access/notebook"
  },
  "source": "meta"
}