{
  "id": 127223,
  "title": "BERT,RoBERTa,ALBERT and...What pretrained model have you used?",
  "url": "/competitions/tensorflow2-question-answering/discussion/127223",
  "author_name": "",
  "post_date": "2020-01-23T00:55:17.133320800Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Because of the kernel running time limit, I have only used a single model, which is RoBERTa-large, I got 0.67.</p>",
  "messages": [
    {
      "id": "726304",
      "postDate": "01/23/2020 00:55:17",
      "content": "<p>Because of the kernel running time limit, I have only used a single model, which is RoBERTa-large, I got 0.67.</p>",
      "rawMarkdown": "Because of the kernel running time limit, I have only used a single model, which is RoBERTa-large, I got 0.67.",
      "votes": null
    },
    {
      "id": "726312",
      "postDate": "01/23/2020 01:00:59",
      "content": "<p>a wierd Roberta-large in keras with squad pretrained by myself, got 0.65</p>",
      "rawMarkdown": "a wierd Roberta-large in keras with squad pretrained by myself, got 0.65",
      "votes": null
    },
    {
      "id": "726331",
      "postDate": "01/23/2020 01:14:08",
      "content": "<p>I have 2 submissions got 0.68 in public LB, but 0.65 in private LB.</p>",
      "rawMarkdown": "I have 2 submissions got 0.68 in public LB, but 0.65 in private LB.",
      "votes": null
    },
    {
      "id": "726772",
      "postDate": "01/23/2020 08:28:26",
      "content": "<p>Roberta-large in pytorch with squad pretrained by myself, got 0.69 private LB</p>",
      "rawMarkdown": "Roberta-large in pytorch with squad pretrained by myself, got 0.69 private LB",
      "votes": null
    },
    {
      "id": "726931",
      "postDate": "01/23/2020 10:18:23",
      "content": "<p><a href=\"/christofhenkel\">@christofhenkel</a> <a href=\"/guozhiyu0914\">@guozhiyu0914</a> - Just curious to know, any other post-processing to select short and long spans other than bert-joint?</p>",
      "rawMarkdown": "christofhenkel @guozhiyu0914 - Just curious to know, any other post-processing to select short and long spans other than bert-joint?",
      "votes": null
    },
    {
      "id": "883198",
      "postDate": "06/12/2020 12:50:59",
      "content": "<p>As i seen that RoBERTa model takes a long running time. so rather than RoBERTa and Bert which model would you suggest to be used for fast running rather than that?\nand can you also share their approx running time!!</p>",
      "rawMarkdown": "As i seen that RoBERTa model takes a long running time. so rather than RoBERTa and Bert which model would you suggest to be used for fast running rather than that?\nand can you also share their approx running time!!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 726312,
      "author_name": "httpwwwfszyc",
      "author_url": "",
      "post_date": "01/23/2020 01:00:59",
      "content": "<p>a wierd Roberta-large in keras with squad pretrained by myself, got 0.65</p>",
      "votes": null,
      "replies": [
        {
          "id": 726331,
          "author_name": "guozhiyu0914",
          "author_url": "",
          "post_date": "01/23/2020 01:14:08",
          "content": "<p>I have 2 submissions got 0.68 in public LB, but 0.65 in private LB.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 726772,
          "author_name": "christofhenkel",
          "author_url": "",
          "post_date": "01/23/2020 08:28:26",
          "content": "<p>Roberta-large in pytorch with squad pretrained by myself, got 0.69 private LB</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 726931,
          "author_name": "s4sarath",
          "author_url": "",
          "post_date": "01/23/2020 10:18:23",
          "content": "<p><a href=\"/christofhenkel\">@christofhenkel</a> <a href=\"/guozhiyu0914\">@guozhiyu0914</a> - Just curious to know, any other post-processing to select short and long spans other than bert-joint?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 883198,
      "author_name": "pranaydate",
      "author_url": "",
      "post_date": "06/12/2020 12:50:59",
      "content": "<p>As i seen that RoBERTa model takes a long running time. so rather than RoBERTa and Bert which model would you suggest to be used for fast running rather than that?\nand can you also share their approx running time!!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "726304": "Because of the kernel running time limit, I have only used a single model, which is RoBERTa-large, I got 0.67.",
    "726312": "a wierd Roberta-large in keras with squad pretrained by myself, got 0.65",
    "726331": "I have 2 submissions got 0.68 in public LB, but 0.65 in private LB.",
    "726772": "Roberta-large in pytorch with squad pretrained by myself, got 0.69 private LB",
    "726931": "christofhenkel @guozhiyu0914 - Just curious to know, any other post-processing to select short and long spans other than bert-joint?",
    "883198": "As i seen that RoBERTa model takes a long running time. so rather than RoBERTa and Bert which model would you suggest to be used for fast running rather than that?\nand can you also share their approx running time!!"
  },
  "source": "meta"
}