{
  "id": 255454,
  "title": "[96%] Fine-tuning ResNet34 with PyTorch",
  "url": "/competitions/cifar-10/discussion/255454",
  "author_name": "FrancescoLorenzo",
  "post_date": "2021-07-27T15:21:51.090000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi everyone,<br>\nI wrote a comprehensive kernel where I show how to leverage Torchvision's pre-trained models to achieve close to state-of-the-art accuracy (96%) on CIFAR-10 with just 20 epochs and no hyperparameter search.</p>\n<p>The kernel is available <strong><a href=\"https://www.kaggle.com/francescolorenzo/96-fine-tuning-resnet34-with-pytorch\" target=\"_blank\">here</a></strong>.</p>\n<p>Let me know if you have any comments or would like me to expand on certain aspects!</p>",
  "messages": [
    {
      "id": 1401809,
      "postDate": "2021-07-27T15:21:51.090Z",
      "content": "<p>Hi everyone,<br>\nI wrote a comprehensive kernel where I show how to leverage Torchvision's pre-trained models to achieve close to state-of-the-art accuracy (96%) on CIFAR-10 with just 20 epochs and no hyperparameter search.</p>\n<p>The kernel is available <strong><a href=\"https://www.kaggle.com/francescolorenzo/96-fine-tuning-resnet34-with-pytorch\" target=\"_blank\">here</a></strong>.</p>\n<p>Let me know if you have any comments or would like me to expand on certain aspects!</p>",
      "rawMarkdown": "Hi everyone,\nI wrote a comprehensive kernel where I show how to leverage Torchvision's pre-trained models to achieve close to state-of-the-art accuracy (96%) on CIFAR-10 with just 20 epochs and no hyperparameter search.\n\nThe kernel is available **[here](https://www.kaggle.com/francescolorenzo/96-fine-tuning-resnet34-with-pytorch)**.\n\nLet me know if you have any comments or would like me to expand on certain aspects!",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1401809": "Hi everyone,\nI wrote a comprehensive kernel where I show how to leverage Torchvision's pre-trained models to achieve close to state-of-the-art accuracy (96%) on CIFAR-10 with just 20 epochs and no hyperparameter search.\n\nThe kernel is available **[here](https://www.kaggle.com/francescolorenzo/96-fine-tuning-resnet34-with-pytorch)**.\n\nLet me know if you have any comments or would like me to expand on certain aspects!"
  }
}