{
  "id": 210152,
  "title": "[Experimental] Different Learning Rate Schedulers",
  "url": "/competitions/rfcx-species-audio-detection/discussion/210152",
  "author_name": "Sinan Calisir",
  "post_date": "2021-01-09T20:41:55.829000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi everyone,</p>\n<p>I was wondering that how the different learning rate schemes would work with this competition in terms of optimization. And, I started this by investigating the different options available. I want to share the initial version of my <a href=\"https://www.kaggle.com/snnclsr/learning-rate-schedulers\" target=\"_blank\">Learning-Rate Schedulers</a> notebook which demonstrates the behaviours of different scheduling algorithms exist in Pytorch and Transformers libraries. I hope to update this post as I try new schedulers. All of my experiments were using <code>CosineAnnealingLR</code> so far ✌️</p>",
  "messages": [
    {
      "id": 1146538,
      "postDate": "2021-01-09T20:41:55.830Z",
      "content": "<p>Hi everyone,</p>\n<p>I was wondering that how the different learning rate schemes would work with this competition in terms of optimization. And, I started this by investigating the different options available. I want to share the initial version of my <a href=\"https://www.kaggle.com/snnclsr/learning-rate-schedulers\" target=\"_blank\">Learning-Rate Schedulers</a> notebook which demonstrates the behaviours of different scheduling algorithms exist in Pytorch and Transformers libraries. I hope to update this post as I try new schedulers. All of my experiments were using <code>CosineAnnealingLR</code> so far ✌️</p>",
      "rawMarkdown": "Hi everyone,\n\nI was wondering that how the different learning rate schemes would work with this competition in terms of optimization. And, I started this by investigating the different options available. I want to share the initial version of my [Learning-Rate Schedulers](https://www.kaggle.com/snnclsr/learning-rate-schedulers) notebook which demonstrates the behaviours of different scheduling algorithms exist in Pytorch and Transformers libraries. I hope to update this post as I try new schedulers. All of my experiments were using `CosineAnnealingLR` so far ✌️",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1146538": "Hi everyone,\n\nI was wondering that how the different learning rate schemes would work with this competition in terms of optimization. And, I started this by investigating the different options available. I want to share the initial version of my [Learning-Rate Schedulers](https://www.kaggle.com/snnclsr/learning-rate-schedulers) notebook which demonstrates the behaviours of different scheduling algorithms exist in Pytorch and Transformers libraries. I hope to update this post as I try new schedulers. All of my experiments were using `CosineAnnealingLR` so far ✌️"
  }
}