{
  "id": 288209,
  "title": "57th Place - Increasing batch size helps ",
  "content": "<p>It is my first Kaggle competition (though I have joined Kaggle 4 years ago 😣). And, thanks to this community.  😍</p>\n<p>I was really lost hope when I  have seen that my ranking in public LB has been decreasing day by day. So, it was a huge jump from 557th to 57th place in the final standings.</p>\n<p>I have tried the following models:</p>\n<ul>\n<li>xlm roberta squad2</li>\n<li>Muril</li>\n</ul>\n<p>In my case, Muril did not perform well on the public leaderboard. So, I have tried to use only Xlm Roberta squad2.</p>\n<p>While increasing batch size from 4 to 16, I have found it performs well. Also, validation loss is not a good measurement to pick the best model in each fold, so I have used Jaccard score.</p>\n<p>Also, when I got stuck, I finetuned my current best model with all external datasets available in Kaggle. </p>\n<p>I have started this competition 24 days ago. I wish I could experiment with blending the models, using other (Rembert, etc) models, and increasing the number of folds.</p>\n<p>Also, another thing, I wish that I could select one of my other models for the final score because this model really performs well in the private score.</p>\n<p>Thanks, <a href=\"https://www.kaggle.com/kishalmandal\" target=\"_blank\">@kishalmandal</a> because of his post-processing and infer code. It really helps me a lot.</p>",
  "comments": [
    {
      "id": 1585474,
      "author_name": "Kishal Mandal",
      "author_url": "",
      "post_date": "2021-11-17T09:18:38.090000",
      "content": "<p>Great job. Best of luck for upcoming competitions :)</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1585714,
          "author_name": "Townim Faisal",
          "author_url": "",
          "post_date": "2021-11-17T13:15:49.067000",
          "content": "<p>Thank you again for your notebook</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "comment_count": 2,
  "messages": [
    {
      "id": 1585215,
      "postDate": "2021-11-17T07:18:25.530Z",
      "content": "<p>It is my first Kaggle competition (though I have joined Kaggle 4 years ago 😣). And, thanks to this community.  😍</p>\n<p>I was really lost hope when I  have seen that my ranking in public LB has been decreasing day by day. So, it was a huge jump from 557th to 57th place in the final standings.</p>\n<p>I have tried the following models:</p>\n<ul>\n<li>xlm roberta squad2</li>\n<li>Muril</li>\n</ul>\n<p>In my case, Muril did not perform well on the public leaderboard. So, I have tried to use only Xlm Roberta squad2.</p>\n<p>While increasing batch size from 4 to 16, I have found it performs well. Also, validation loss is not a good measurement to pick the best model in each fold, so I have used Jaccard score.</p>\n<p>Also, when I got stuck, I finetuned my current best model with all external datasets available in Kaggle. </p>\n<p>I have started this competition 24 days ago. I wish I could experiment with blending the models, using other (Rembert, etc) models, and increasing the number of folds.</p>\n<p>Also, another thing, I wish that I could select one of my other models for the final score because this model really performs well in the private score.</p>\n<p>Thanks, <a href=\"https://www.kaggle.com/kishalmandal\" target=\"_blank\">@kishalmandal</a> because of his post-processing and infer code. It really helps me a lot.</p>",
      "rawMarkdown": "It is my first Kaggle competition (though I have joined Kaggle 4 years ago 😣). And, thanks to this community.  😍\n\nI was really lost hope when I  have seen that my ranking in public LB has been decreasing day by day. So, it was a huge jump from 557th to 57th place in the final standings.\n\nI have tried the following models:\n- xlm roberta squad2\n- Muril\n\nIn my case, Muril did not perform well on the public leaderboard. So, I have tried to use only Xlm Roberta squad2.\n\nWhile increasing batch size from 4 to 16, I have found it performs well. Also, validation loss is not a good measurement to pick the best model in each fold, so I have used Jaccard score.\n\nAlso, when I got stuck, I finetuned my current best model with all external datasets available in Kaggle. \n\nI have started this competition 24 days ago. I wish I could experiment with blending the models, using other (Rembert, etc) models, and increasing the number of folds.\n\nAlso, another thing, I wish that I could select one of my other models for the final score because this model really performs well in the private score.\n\nThanks, @kishalmandal because of his post-processing and infer code. It really helps me a lot.",
      "votes": 9
    },
    {
      "id": 1585474,
      "postDate": "2021-11-17T09:18:38.090Z",
      "content": "<p>Great job. Best of luck for upcoming competitions :)</p>",
      "rawMarkdown": "Great job. Best of luck for upcoming competitions :)",
      "votes": 1,
      "replies": [
        {
          "id": 1585714,
          "postDate": "2021-11-17T13:15:49.067Z",
          "content": "<p>Thank you again for your notebook</p>",
          "rawMarkdown": "Thank you again for your notebook"
        }
      ]
    }
  ],
  "complete": true,
  "source": "live",
  "fetched_at": "2026-10-19T16:18:59Z",
  "competition_id": 30060,
  "competition_slug": "chaii-hindi-and-tamil-question-answering",
  "team_id": 7580188,
  "solution_url": "/competitions/chaii-hindi-and-tamil-question-answering/writeups/bsds-57th-place-increasing-batch-size-helps"
}