{
  "id": 261982,
  "title": "What about using different optimizers?",
  "url": "/competitions/seti-breakthrough-listen/discussion/261982",
  "author_name": "Aristarkh_BFG",
  "post_date": "2021-08-05T13:31:48.015000",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Most likely, many of us use standard optimizers, like Adam. But I would like to know if you have experimented with different optimizers? I attach a link to an excellent visualization for Pytorch. <a href=\"https://github.com/jettify/pytorch-optimizer\" target=\"_blank\">https://github.com/jettify/pytorch-optimizer</a></p>",
  "messages": [
    {
      "id": 1451939,
      "postDate": "2021-08-05T13:31:48.017Z",
      "content": "<p>Most likely, many of us use standard optimizers, like Adam. But I would like to know if you have experimented with different optimizers? I attach a link to an excellent visualization for Pytorch. <a href=\"https://github.com/jettify/pytorch-optimizer\" target=\"_blank\">https://github.com/jettify/pytorch-optimizer</a></p>",
      "rawMarkdown": "Most likely, many of us use standard optimizers, like Adam. But I would like to know if you have experimented with different optimizers? I attach a link to an excellent visualization for Pytorch. https://github.com/jettify/pytorch-optimizer",
      "votes": 4
    },
    {
      "id": 1459883,
      "postDate": "2021-08-08T15:38:53.617Z",
      "content": "<p>try ranger optimizer (combination of Lookahead + Radam) works for me. Ranger will converge faster than Adam.</p>",
      "rawMarkdown": "try ranger optimizer (combination of Lookahead + Radam) works for me. Ranger will converge faster than Adam."
    },
    {
      "id": 1452626,
      "postDate": "2021-08-05T16:47:37.013Z",
      "content": "<p>In most cases Adam or AdamW is more than sufficient for model training. A data preprocessing and a model structure are much more important.</p>",
      "rawMarkdown": "In most cases Adam or AdamW is more than sufficient for model training. A data preprocessing and a model structure are much more important."
    }
  ],
  "comments": [
    {
      "id": 1459883,
      "author_name": "Parth Dhameliya",
      "author_url": "",
      "post_date": "2021-08-08T15:38:53.617000",
      "content": "<p>try ranger optimizer (combination of Lookahead + Radam) works for me. Ranger will converge faster than Adam.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1452626,
      "author_name": "Araik Tamazian",
      "author_url": "",
      "post_date": "2021-08-05T16:47:37.013000",
      "content": "<p>In most cases Adam or AdamW is more than sufficient for model training. A data preprocessing and a model structure are much more important.</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1451939": "Most likely, many of us use standard optimizers, like Adam. But I would like to know if you have experimented with different optimizers? I attach a link to an excellent visualization for Pytorch. https://github.com/jettify/pytorch-optimizer",
    "1459883": "try ranger optimizer (combination of Lookahead + Radam) works for me. Ranger will converge faster than Adam.",
    "1452626": "In most cases Adam or AdamW is more than sufficient for model training. A data preprocessing and a model structure are much more important."
  }
}