{
  "id": 201223,
  "title": "[Models] -  incremental training for models",
  "url": "/competitions/riiid-test-answer-prediction/discussion/201223",
  "author_name": "Thamer",
  "post_date": "2020-12-03T18:34:26.469000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi , <br>\nIt is a good idea to test incremental training for boost models (xgb and lgbm) using epochs ? see exemple below : </p>\n<pre><code>params = {'objective': 'reg:linear', 'verbose': False}\nmodel_1 = xgb.train(params, xg_train_1, 30)\nmodel_1.save_model('model_1.model')\n\n# ================= train two versions of the model =====================#\nmodel_2_v1 = xgb.train(params, xg_train_2, 30)\nmodel_2_v2 = xgb.train(params, xg_train_2, 30, xgb_model='model_1.model')\n</code></pre>\n<p>PS: Sorry to ask but i dont have time to test it: i joined the challenge yesterday :'(</p>",
  "messages": [
    {
      "id": 1101247,
      "postDate": "2020-12-03T18:34:26.470Z",
      "content": "<p>Hi , <br>\nIt is a good idea to test incremental training for boost models (xgb and lgbm) using epochs ? see exemple below : </p>\n<pre><code>params = {'objective': 'reg:linear', 'verbose': False}\nmodel_1 = xgb.train(params, xg_train_1, 30)\nmodel_1.save_model('model_1.model')\n\n# ================= train two versions of the model =====================#\nmodel_2_v1 = xgb.train(params, xg_train_2, 30)\nmodel_2_v2 = xgb.train(params, xg_train_2, 30, xgb_model='model_1.model')\n</code></pre>\n<p>PS: Sorry to ask but i dont have time to test it: i joined the challenge yesterday :'(</p>",
      "rawMarkdown": "Hi , \nIt is a good idea to test incremental training for boost models (xgb and lgbm) using epochs ? see exemple below : \n```\nparams = {'objective': 'reg:linear', 'verbose': False}\nmodel_1 = xgb.train(params, xg_train_1, 30)\nmodel_1.save_model('model_1.model')\n\n# ================= train two versions of the model =====================#\nmodel_2_v1 = xgb.train(params, xg_train_2, 30)\nmodel_2_v2 = xgb.train(params, xg_train_2, 30, xgb_model='model_1.model')\n\n```\n\n\nPS: Sorry to ask but i dont have time to test it: i joined the challenge yesterday :'("
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1101247": "Hi , \nIt is a good idea to test incremental training for boost models (xgb and lgbm) using epochs ? see exemple below : \n```\nparams = {'objective': 'reg:linear', 'verbose': False}\nmodel_1 = xgb.train(params, xg_train_1, 30)\nmodel_1.save_model('model_1.model')\n\n# ================= train two versions of the model =====================#\nmodel_2_v1 = xgb.train(params, xg_train_2, 30)\nmodel_2_v2 = xgb.train(params, xg_train_2, 30, xgb_model='model_1.model')\n\n```\n\n\nPS: Sorry to ask but i dont have time to test it: i joined the challenge yesterday :'("
  }
}