{
  "id": 237121,
  "title": "A simple and clean pytorch-lightning pipeline",
  "url": "/competitions/birdclef-2021/discussion/237121",
  "author_name": "27Wilson",
  "post_date": "2021-05-07T08:33:26.749000",
  "votes": 7,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hello everyone, I'm 27Wilson and I just joined this competition. I'm new to deep learning, and I want to learn some skills from kaggle, any suggestions are welcome. <br>\nI find <a href=\"https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-training-colab\" target=\"_blank\">kenroma's notebook</a> is easy to start, but the workflow is a little comfused to me. So I public a kernel <a href=\"https://www.kaggle.com/whurobin/training-pipeline-in-pytorch-lightning\" target=\"_blank\">here</a> which use pytorch-lightning to control workflow and timm to load pretrained model. You can also use <a href=\"https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-inference\" target=\"_blank\">this kernel</a> to do inference, just change a little code.<br>\nI will try to use bigger models, augmentation tircks, and so on.</p>",
  "messages": [
    {
      "id": 1296393,
      "postDate": "2021-05-07T08:33:26.750Z",
      "content": "<p>Hello everyone, I'm 27Wilson and I just joined this competition. I'm new to deep learning, and I want to learn some skills from kaggle, any suggestions are welcome. <br>\nI find <a href=\"https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-training-colab\" target=\"_blank\">kenroma's notebook</a> is easy to start, but the workflow is a little comfused to me. So I public a kernel <a href=\"https://www.kaggle.com/whurobin/training-pipeline-in-pytorch-lightning\" target=\"_blank\">here</a> which use pytorch-lightning to control workflow and timm to load pretrained model. You can also use <a href=\"https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-inference\" target=\"_blank\">this kernel</a> to do inference, just change a little code.<br>\nI will try to use bigger models, augmentation tircks, and so on.</p>",
      "rawMarkdown": "Hello everyone, I'm 27Wilson and I just joined this competition. I'm new to deep learning, and I want to learn some skills from kaggle, any suggestions are welcome. \nI find [kenroma's notebook](https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-training-colab) is easy to start, but the workflow is a little comfused to me. So I public a kernel [here](https://www.kaggle.com/whurobin/training-pipeline-in-pytorch-lightning) which use pytorch-lightning to control workflow and timm to load pretrained model. You can also use [this kernel](https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-inference) to do inference, just change a little code.\nI will try to use bigger models, augmentation tircks, and so on.",
      "votes": 7
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1296393": "Hello everyone, I'm 27Wilson and I just joined this competition. I'm new to deep learning, and I want to learn some skills from kaggle, any suggestions are welcome. \nI find [kenroma's notebook](https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-training-colab) is easy to start, but the workflow is a little comfused to me. So I public a kernel [here](https://www.kaggle.com/whurobin/training-pipeline-in-pytorch-lightning) which use pytorch-lightning to control workflow and timm to load pretrained model. You can also use [this kernel](https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-inference) to do inference, just change a little code.\nI will try to use bigger models, augmentation tircks, and so on."
  }
}