{
  "id": 115148,
  "title": "Getting Too odd lb score",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/115148",
  "author_name": "Jaideep",
  "post_date": "2019-10-31T16:37:25.831000",
  "votes": 0,
  "comment_count": 4,
  "views": 0,
  "content": "<p>My cv loss is just 0.058\nbut lb score is 0.33\nHas any one faced this issue before.\nWhat mistake i could be making during inference..</p>",
  "messages": [
    {
      "id": 662531,
      "postDate": "2019-10-31T16:55:11.083Z",
      "content": "<p>classes order is not consistent (any is first instead of last)?</p>",
      "rawMarkdown": "classes order is not consistent (any is first instead of last)?",
      "replies": [
        {
          "id": 663284,
          "postDate": "2019-11-01T18:26:59.223Z",
          "content": "<p><a href=\"/zaharch\">@zaharch</a> \ni see that in sample submission any comes in last but many of us while buidling df for train put any as class 0 so label 0 out of 6 corresponds to any ..\nis that what you are saying ?</p>",
          "rawMarkdown": "@zaharch \ni see that in sample submission any comes in last but many of us while buidling df for train put any as class 0 so label 0 out of 6 corresponds to any ..\nis that what you are saying ?\n"
        },
        {
          "id": 663292,
          "postDate": "2019-11-01T18:47:57.647Z",
          "content": "<p>Yes, but for submission file it does not matter, can be any order. But if in your loss function you assume one order, and for the test predictions another order, - this is a problem.</p>",
          "rawMarkdown": "Yes, but for submission file it does not matter, can be any order. But if in your loss function you assume one order, and for the test predictions another order, - this is a problem."
        }
      ]
    },
    {
      "id": 662517,
      "postDate": "2019-10-31T16:37:25.830Z",
      "content": "<p>My cv loss is just 0.058\nbut lb score is 0.33\nHas any one faced this issue before.\nWhat mistake i could be making during inference..</p>",
      "rawMarkdown": "My cv loss is just 0.058\nbut lb score is 0.33\nHas any one faced this issue before.\nWhat mistake i could be making during inference..\n"
    },
    {
      "id": 663124,
      "postDate": "2019-11-01T13:34:07.443Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 662531,
      "author_name": "nosound",
      "author_url": "",
      "post_date": "2019-10-31T16:55:11.083000",
      "content": "<p>classes order is not consistent (any is first instead of last)?</p>",
      "votes": 0,
      "replies": [
        {
          "id": 663284,
          "author_name": "Jaideep",
          "author_url": "",
          "post_date": "2019-11-01T18:26:59.223000",
          "content": "<p><a href=\"/zaharch\">@zaharch</a> \ni see that in sample submission any comes in last but many of us while buidling df for train put any as class 0 so label 0 out of 6 corresponds to any ..\nis that what you are saying ?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 663292,
          "author_name": "nosound",
          "author_url": "",
          "post_date": "2019-11-01T18:47:57.647000",
          "content": "<p>Yes, but for submission file it does not matter, can be any order. But if in your loss function you assume one order, and for the test predictions another order, - this is a problem.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 663124,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-11-01T13:34:07.443000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "662531": "classes order is not consistent (any is first instead of last)?",
    "662517": "My cv loss is just 0.058\nbut lb score is 0.33\nHas any one faced this issue before.\nWhat mistake i could be making during inference..\n",
    "663124": ""
  }
}