{
  "id": 186256,
  "title": "Better optimization: AdamP",
  "url": "/competitions/lyft-motion-prediction-autonomous-vehicles/discussion/186256",
  "author_name": "",
  "post_date": "2020-09-23T18:45:52.034624300Z",
  "votes": 32,
  "comment_count": 6,
  "views": 0,
  "content": "<p>I advice to use AdamP. I have good result with this optimization. It help my loss more smoth and coverge quickly. <br>\n<a href=\"https://github.com/clovaai/AdamP\" target=\"_blank\">https://github.com/clovaai/AdamP</a></p>\n<pre><code>from adamp import AdamP\noptimizer = AdamP(params, lr=0.0001, betas=(0.9, 0.999), weight_decay=1e-2)\n</code></pre>",
  "messages": [
    {
      "id": "1024318",
      "postDate": "09/23/2020 18:45:52",
      "content": "<p>I advice to use AdamP. I have good result with this optimization. It help my loss more smoth and coverge quickly. <br>\n<a href=\"https://github.com/clovaai/AdamP\" target=\"_blank\">https://github.com/clovaai/AdamP</a></p>\n<pre><code>from adamp import AdamP\noptimizer = AdamP(params, lr=0.0001, betas=(0.9, 0.999), weight_decay=1e-2)\n</code></pre>",
      "rawMarkdown": "I advice to use AdamP. I have good result with this optimization. It help my loss more smoth and coverge quickly. \nhttps://github.com/clovaai/AdamP\n```\nfrom adamp import AdamP\noptimizer = AdamP(params, lr=0.0001, betas=(0.9, 0.999), weight_decay=1e-2)\n```",
      "votes": null
    },
    {
      "id": "1025724",
      "postDate": "09/24/2020 18:19:54",
      "content": "<p>Informative , Thanks for sharing </p>",
      "rawMarkdown": "Informative , Thanks for sharing",
      "votes": null
    },
    {
      "id": "1027364",
      "postDate": "09/26/2020 05:18:31",
      "content": "<p>You can also try to tune hyperparameters in it to get more refined results.</p>",
      "rawMarkdown": "You can also try to tune hyperparameters in it to get more refined results.",
      "votes": null
    },
    {
      "id": "1028422",
      "postDate": "09/26/2020 20:43:00",
      "content": "<p>Thank you, I will give it a try</p>",
      "rawMarkdown": "Thank you, I will give it a try",
      "votes": null
    },
    {
      "id": "1029110",
      "postDate": "09/27/2020 13:17:27",
      "content": "<p>Thank you for this!</p>",
      "rawMarkdown": "Thank you for this!",
      "votes": null
    },
    {
      "id": "1029919",
      "postDate": "09/28/2020 09:00:43",
      "content": "<p>May not be true for all scenarios. Use various types of optimizers based on the needs and after performing hyperpameter tuning.</p>",
      "rawMarkdown": "May not be true for all scenarios. Use various types of optimizers based on the needs and after performing hyperpameter tuning.",
      "votes": null
    },
    {
      "id": "1032041",
      "postDate": "09/29/2020 22:26:13",
      "content": "<p>Did anyone try it and got better results? I'm \"fighting\" with some <code>NaN</code> outputs right now…</p>",
      "rawMarkdown": "Did anyone try it and got better results? I'm \"fighting\" with some `NaN` outputs right now...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1027364,
      "author_name": "shivambhardwaj0101",
      "author_url": "",
      "post_date": "09/26/2020 05:18:31",
      "content": "<p>You can also try to tune hyperparameters in it to get more refined results.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1028422,
      "author_name": "aliabdin1",
      "author_url": "",
      "post_date": "09/26/2020 20:43:00",
      "content": "<p>Thank you, I will give it a try</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1029110,
      "author_name": "fireheart7",
      "author_url": "",
      "post_date": "09/27/2020 13:17:27",
      "content": "<p>Thank you for this!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1029919,
      "author_name": "srinuti",
      "author_url": "",
      "post_date": "09/28/2020 09:00:43",
      "content": "<p>May not be true for all scenarios. Use various types of optimizers based on the needs and after performing hyperpameter tuning.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1032041,
      "author_name": "aliabdin1",
      "author_url": "",
      "post_date": "09/29/2020 22:26:13",
      "content": "<p>Did anyone try it and got better results? I'm \"fighting\" with some <code>NaN</code> outputs right now…</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1025724,
      "author_name": "pinakimishrads",
      "author_url": "",
      "post_date": "09/24/2020 18:19:54",
      "content": "<p>Informative , Thanks for sharing </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1024318": "I advice to use AdamP. I have good result with this optimization. It help my loss more smoth and coverge quickly. \nhttps://github.com/clovaai/AdamP\n```\nfrom adamp import AdamP\noptimizer = AdamP(params, lr=0.0001, betas=(0.9, 0.999), weight_decay=1e-2)\n```",
    "1025724": "Informative , Thanks for sharing",
    "1027364": "You can also try to tune hyperparameters in it to get more refined results.",
    "1028422": "Thank you, I will give it a try",
    "1029110": "Thank you for this!",
    "1029919": "May not be true for all scenarios. Use various types of optimizers based on the needs and after performing hyperpameter tuning.",
    "1032041": "Did anyone try it and got better results? I'm \"fighting\" with some `NaN` outputs right now..."
  },
  "source": "meta"
}