{
  "id": 94115,
  "title": "Lightgbm num_threads",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/94115",
  "author_name": "",
  "post_date": "2019-06-02T07:54:09.157088500Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi guys, I'm new to Lightgbm and it seems that increasing num_threads can increase training speed. However, will it worsen my score? If so, what is the ideal num_threads i should be using?</p>",
  "messages": [
    {
      "id": "541300",
      "postDate": "06/02/2019 07:54:09",
      "content": "<p>Hi guys, I'm new to Lightgbm and it seems that increasing num_threads can increase training speed. However, will it worsen my score? If so, what is the ideal num_threads i should be using?</p>",
      "rawMarkdown": "Hi guys, I'm new to Lightgbm and it seems that increasing num_threads can increase training speed. However, will it worsen my score? If so, what is the ideal num_threads i should be using?",
      "votes": null
    },
    {
      "id": "541391",
      "postDate": "06/02/2019 11:14:46",
      "content": "<p>I use one thread per cpu, i.e. 4 currently.  With small datasets you can go to one thread per logical cpu, i.e. 2x more threads.  I could therefore go to 8 threads, but the dataset is so small that running time isn't an issue with my 15 or less features.</p>",
      "rawMarkdown": "I use one thread per cpu, i.e. 4 currently.  With small datasets you can go to one thread per logical cpu, i.e. 2x more threads.  I could therefore go to 8 threads, but the dataset is so small that running time isn't an issue with my 15 or less features.",
      "votes": null
    },
    {
      "id": "541509",
      "postDate": "06/02/2019 15:14:13",
      "content": "<p>Got it, thanks!</p>",
      "rawMarkdown": "Got it, thanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 541391,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "06/02/2019 11:14:46",
      "content": "<p>I use one thread per cpu, i.e. 4 currently.  With small datasets you can go to one thread per logical cpu, i.e. 2x more threads.  I could therefore go to 8 threads, but the dataset is so small that running time isn't an issue with my 15 or less features.</p>",
      "votes": null,
      "replies": [
        {
          "id": 541509,
          "author_name": "kwquan",
          "author_url": "",
          "post_date": "06/02/2019 15:14:13",
          "content": "<p>Got it, thanks!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "541300": "Hi guys, I'm new to Lightgbm and it seems that increasing num_threads can increase training speed. However, will it worsen my score? If so, what is the ideal num_threads i should be using?",
    "541391": "I use one thread per cpu, i.e. 4 currently.  With small datasets you can go to one thread per logical cpu, i.e. 2x more threads.  I could therefore go to 8 threads, but the dataset is so small that running time isn't an issue with my 15 or less features.",
    "541509": "Got it, thanks!"
  },
  "source": "meta"
}