{
  "id": 458997,
  "title": "Loading ViT: Vision Transformer",
  "url": "/competitions/UBC-OCEAN/discussion/458997",
  "author_name": "Aditya Parikh",
  "post_date": "2023-12-02T20:58:24.501000",
  "votes": 3,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello, Is it possible to load a vision transformer (base-224) for this competition in offline (internet-off) environment?<br>\nI tried loading it using timm and huggingface transformer but failed. <br>\nCan anyone help?</p>",
  "messages": [
    {
      "id": 2549791,
      "postDate": "2023-12-05T14:24:04.837Z",
      "content": "<p>Do you really need to load it offline, train, and run predictions?<br>\n I would suggest loading the pre-trained model online, training it, and saving the final model which you can add as input to your offline kernel used for submission, see the two-step solution here:</p>\n<ol>\n<li><a href=\"https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm\" target=\"_blank\">https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm</a></li>\n<li><a href=\"https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lit-torch-inference-thumbnails\" target=\"_blank\">https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lit-torch-inference-thumbnails</a></li>\n</ol>",
      "rawMarkdown": "Do you really need to load it offline, train, and run predictions?\n I would suggest loading the pre-trained model online, training it, and saving the final model which you can add as input to your offline kernel used for submission, see the two-step solution here:\n1. https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm\n2. https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lit-torch-inference-thumbnails\n",
      "votes": 3
    },
    {
      "id": 2546869,
      "postDate": "2023-12-03T00:25:01.393Z",
      "content": "<p>The answer to this post on timm's discussion board (<a href=\"url\" target=\"_blank\">https://github.com/huggingface/pytorch-image-models/discussions/1826</a>) gives you the steps to be able to set it up.</p>",
      "rawMarkdown": "The answer to this post on timm's discussion board ([https://github.com/huggingface/pytorch-image-models/discussions/1826](url)) gives you the steps to be able to set it up.",
      "votes": 3
    },
    {
      "id": 2546794,
      "postDate": "2023-12-02T20:58:24.503Z",
      "content": "<p>Hello, Is it possible to load a vision transformer (base-224) for this competition in offline (internet-off) environment?<br>\nI tried loading it using timm and huggingface transformer but failed. <br>\nCan anyone help?</p>",
      "rawMarkdown": "Hello, Is it possible to load a vision transformer (base-224) for this competition in offline (internet-off) environment?\nI tried loading it using timm and huggingface transformer but failed. \nCan anyone help?",
      "votes": 3
    }
  ],
  "comments": [
    {
      "id": 2549791,
      "author_name": "Jirka",
      "author_url": "",
      "post_date": "2023-12-05T14:24:04.837000",
      "content": "<p>Do you really need to load it offline, train, and run predictions?<br>\n I would suggest loading the pre-trained model online, training it, and saving the final model which you can add as input to your offline kernel used for submission, see the two-step solution here:</p>\n<ol>\n<li><a href=\"https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm\" target=\"_blank\">https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm</a></li>\n<li><a href=\"https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lit-torch-inference-thumbnails\" target=\"_blank\">https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lit-torch-inference-thumbnails</a></li>\n</ol>",
      "votes": 3,
      "replies": []
    },
    {
      "id": 2546869,
      "author_name": "Patrick Robitaille",
      "author_url": "",
      "post_date": "2023-12-03T00:25:01.393000",
      "content": "<p>The answer to this post on timm's discussion board (<a href=\"url\" target=\"_blank\">https://github.com/huggingface/pytorch-image-models/discussions/1826</a>) gives you the steps to be able to set it up.</p>",
      "votes": 3,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2549791": "Do you really need to load it offline, train, and run predictions?\n I would suggest loading the pre-trained model online, training it, and saving the final model which you can add as input to your offline kernel used for submission, see the two-step solution here:\n1. https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm\n2. https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lit-torch-inference-thumbnails\n",
    "2546869": "The answer to this post on timm's discussion board ([https://github.com/huggingface/pytorch-image-models/discussions/1826](url)) gives you the steps to be able to set it up.",
    "2546794": "Hello, Is it possible to load a vision transformer (base-224) for this competition in offline (internet-off) environment?\nI tried loading it using timm and huggingface transformer but failed. \nCan anyone help?"
  }
}