{
  "id": 508269,
  "title": "Continuous format is not supported(Metrics)",
  "url": "/competitions/birdclef-2024/discussion/508269",
  "author_name": "",
  "post_date": "2024-05-28T23:50:16.336749900Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Dear Kagglers !<br>\nI'm poor at English, so if you have any vague points, don't hesitate to ask me for more details !</p>\n<p>When I tried to score my prediction, I got the error \"sklearn.metric.roc_auc does not support continuous format.\"<br>\nI think this is caused by mix-up augmentation.<br>\n(e.g. [1, 0] * 0.3 + [0, 1] * 0.7 = [0.3, 0.7] ← this is continuous format, I think.)</p>\n<p>Then, what to do to solve that error?</p>\n<p>Please share any idea with me!</p>",
  "messages": [
    {
      "id": "2842183",
      "postDate": "05/28/2024 23:50:16",
      "content": "<p>Dear Kagglers !<br>\nI'm poor at English, so if you have any vague points, don't hesitate to ask me for more details !</p>\n<p>When I tried to score my prediction, I got the error \"sklearn.metric.roc_auc does not support continuous format.\"<br>\nI think this is caused by mix-up augmentation.<br>\n(e.g. [1, 0] * 0.3 + [0, 1] * 0.7 = [0.3, 0.7] ← this is continuous format, I think.)</p>\n<p>Then, what to do to solve that error?</p>\n<p>Please share any idea with me!</p>",
      "rawMarkdown": "Dear Kagglers !\nI'm poor at English, so if you have any vague points, don't hesitate to ask me for more details !\n\nWhen I tried to score my prediction, I got the error \"sklearn.metric.roc_auc does not support continuous format.\"\nI think this is caused by mix-up augmentation.\n(e.g. [1, 0] * 0.3 + [0, 1] * 0.7 = [0.3, 0.7] ← this is continuous format, I think.)\n\nThen, what to do to solve that error?\n\nPlease share any idea with me!",
      "votes": null
    },
    {
      "id": "2842267",
      "postDate": "05/29/2024 02:17:22",
      "content": "<p>It has nothing to do with augmentation.<br>\nThe error is during your metrics calculation, because roc_auc needs \"discrete\" value (0 and 1 only) and not continuous (0.2,0.3 etc ). <br>\nCheck the ground truth label that you feed into your metrics calculation, make sure they are binary. <br>\nIf they are not binary, you can try rounding them to 0 and 1 to see if the error disappear. </p>",
      "rawMarkdown": "It has nothing to do with augmentation.\nThe error is during your metrics calculation, because roc_auc needs \"discrete\" value (0 and 1 only) and not continuous (0.2,0.3 etc ). \nCheck the ground truth label that you feed into your metrics calculation, make sure they are binary. \nIf they are not binary, you can try rounding them to 0 and 1 to see if the error disappear.",
      "votes": null
    },
    {
      "id": "2842540",
      "postDate": "05/29/2024 06:21:04",
      "content": "<p>I have met the same error before, it caused by two reasons, first is that you may use the secondary label in the dataset, so that the label become uncontinuous format like [0, 1, …., 0, 1, 0] or [0, 1, …, 0, 0.9, 0], second one is you mentioned mix-up augmentation. So you can try not to use secondary label or mix-up when testing.</p>",
      "rawMarkdown": "I have met the same error before, it caused by two reasons, first is that you may use the secondary label in the dataset, so that the label become uncontinuous format like [0, 1, ...., 0, 1, 0] or [0, 1, ..., 0, 0.9, 0], second one is you mentioned mix-up augmentation. So you can try not to use secondary label or mix-up when testing.",
      "votes": null
    },
    {
      "id": "2845026",
      "postDate": "05/30/2024 11:10:22",
      "content": "<p>Thanks to you, the error seemed to be solved!<br>\nThank you! </p>",
      "rawMarkdown": "Thanks to you, the error seemed to be solved!\nThank you!",
      "votes": null
    },
    {
      "id": "2850159",
      "postDate": "06/02/2024 02:06:17",
      "content": "<p>hii, are you still looking for teammate? </p>",
      "rawMarkdown": "hii, are you still looking for teammate?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2842267,
      "author_name": "nyleve",
      "author_url": "",
      "post_date": "05/29/2024 02:17:22",
      "content": "<p>It has nothing to do with augmentation.<br>\nThe error is during your metrics calculation, because roc_auc needs \"discrete\" value (0 and 1 only) and not continuous (0.2,0.3 etc ). <br>\nCheck the ground truth label that you feed into your metrics calculation, make sure they are binary. <br>\nIf they are not binary, you can try rounding them to 0 and 1 to see if the error disappear. </p>",
      "votes": null,
      "replies": [
        {
          "id": 2845026,
          "author_name": "ueight8",
          "author_url": "",
          "post_date": "05/30/2024 11:10:22",
          "content": "<p>Thanks to you, the error seemed to be solved!<br>\nThank you! </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2842540,
      "author_name": "tanxxx",
      "author_url": "",
      "post_date": "05/29/2024 06:21:04",
      "content": "<p>I have met the same error before, it caused by two reasons, first is that you may use the secondary label in the dataset, so that the label become uncontinuous format like [0, 1, …., 0, 1, 0] or [0, 1, …, 0, 0.9, 0], second one is you mentioned mix-up augmentation. So you can try not to use secondary label or mix-up when testing.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2850159,
          "author_name": "daniellee724",
          "author_url": "",
          "post_date": "06/02/2024 02:06:17",
          "content": "<p>hii, are you still looking for teammate? </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2842183": "Dear Kagglers !\nI'm poor at English, so if you have any vague points, don't hesitate to ask me for more details !\n\nWhen I tried to score my prediction, I got the error \"sklearn.metric.roc_auc does not support continuous format.\"\nI think this is caused by mix-up augmentation.\n(e.g. [1, 0] * 0.3 + [0, 1] * 0.7 = [0.3, 0.7] ← this is continuous format, I think.)\n\nThen, what to do to solve that error?\n\nPlease share any idea with me!",
    "2842267": "It has nothing to do with augmentation.\nThe error is during your metrics calculation, because roc_auc needs \"discrete\" value (0 and 1 only) and not continuous (0.2,0.3 etc ). \nCheck the ground truth label that you feed into your metrics calculation, make sure they are binary. \nIf they are not binary, you can try rounding them to 0 and 1 to see if the error disappear.",
    "2842540": "I have met the same error before, it caused by two reasons, first is that you may use the secondary label in the dataset, so that the label become uncontinuous format like [0, 1, ...., 0, 1, 0] or [0, 1, ..., 0, 0.9, 0], second one is you mentioned mix-up augmentation. So you can try not to use secondary label or mix-up when testing.",
    "2845026": "Thanks to you, the error seemed to be solved!\nThank you!",
    "2850159": "hii, are you still looking for teammate?"
  },
  "source": "meta"
}