{
  "id": 135321,
  "title": "About K-Fold in Pytorch with images",
  "url": "/competitions/bengaliai-cv19/discussion/135321",
  "author_name": "",
  "post_date": "2020-03-13T07:30:35.794437300Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi, everyone, I'm using pytorch to do this competiton, I find many people used k-fold cv to eval models, but I don't know how to do with pytorch, May I need to change NN Frame to Keras of Fastai? </p>",
  "messages": [
    {
      "id": "770639",
      "postDate": "03/13/2020 07:30:35",
      "content": "<p>Hi, everyone, I'm using pytorch to do this competiton, I find many people used k-fold cv to eval models, but I don't know how to do with pytorch, May I need to change NN Frame to Keras of Fastai? </p>",
      "rawMarkdown": "Hi, everyone, I'm using pytorch to do this competiton, I find many people used k-fold cv to eval models, but I don't know how to do with pytorch, May I need to change NN Frame to Keras of Fastai?",
      "votes": null
    },
    {
      "id": "770854",
      "postDate": "03/13/2020 13:36:15",
      "content": "<p>It is just splitting the data into folds, you don't need a framework for this. Split the data into 5 equal folds (with similar distribution will be recommended) and then train on 4 of them and validate on the 5th, after this just permute the folds</p>",
      "rawMarkdown": "It is just splitting the data into folds, you don't need a framework for this. Split the data into 5 equal folds (with similar distribution will be recommended) and then train on 4 of them and validate on the 5th, after this just permute the folds",
      "votes": null
    },
    {
      "id": "770965",
      "postDate": "03/13/2020 16:05:30",
      "content": "<p>Thank u, I'll try it.</p>",
      "rawMarkdown": "Thank u, I'll try it.",
      "votes": null
    },
    {
      "id": "771230",
      "postDate": "03/13/2020 22:26:02",
      "content": "<p>Check out this notebook:\n<a href=\"https://www.kaggle.com/yiheng/iterative-stratification\">https://www.kaggle.com/yiheng/iterative-stratification</a>\nI found it very useful</p>",
      "rawMarkdown": "Check out this notebook:\nhttps://www.kaggle.com/yiheng/iterative-stratification\nI found it very useful",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 770854,
      "author_name": "vladvdv",
      "author_url": "",
      "post_date": "03/13/2020 13:36:15",
      "content": "<p>It is just splitting the data into folds, you don't need a framework for this. Split the data into 5 equal folds (with similar distribution will be recommended) and then train on 4 of them and validate on the 5th, after this just permute the folds</p>",
      "votes": null,
      "replies": [
        {
          "id": 770965,
          "author_name": "fitz007",
          "author_url": "",
          "post_date": "03/13/2020 16:05:30",
          "content": "<p>Thank u, I'll try it.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 771230,
      "author_name": "yuanlin08",
      "author_url": "",
      "post_date": "03/13/2020 22:26:02",
      "content": "<p>Check out this notebook:\n<a href=\"https://www.kaggle.com/yiheng/iterative-stratification\">https://www.kaggle.com/yiheng/iterative-stratification</a>\nI found it very useful</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "770639": "Hi, everyone, I'm using pytorch to do this competiton, I find many people used k-fold cv to eval models, but I don't know how to do with pytorch, May I need to change NN Frame to Keras of Fastai?",
    "770854": "It is just splitting the data into folds, you don't need a framework for this. Split the data into 5 equal folds (with similar distribution will be recommended) and then train on 4 of them and validate on the 5th, after this just permute the folds",
    "770965": "Thank u, I'll try it.",
    "771230": "Check out this notebook:\nhttps://www.kaggle.com/yiheng/iterative-stratification\nI found it very useful"
  },
  "source": "meta"
}