{
  "id": 137935,
  "title": "Notebooks that gave 90+ score now give a score of around 55; Edit: issue resolved; user error",
  "url": "/competitions/flower-classification-with-tpus/discussion/137935",
  "author_name": "",
  "post_date": "2020-03-23T05:05:22.955053Z",
  "votes": null,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Just what the title says...\nAnyone else facing this..?</p>",
  "messages": [
    {
      "id": "783231",
      "postDate": "03/23/2020 05:05:22",
      "content": "<p>Just what the title says...\nAnyone else facing this..?</p>",
      "rawMarkdown": "Just what the title says...\nAnyone else facing this..?",
      "votes": null
    },
    {
      "id": "784056",
      "postDate": "03/23/2020 23:00:11",
      "content": "<p>This looks like a bad case of overfitting... Do you suspect something else like a bug ?</p>",
      "rawMarkdown": "This looks like a bad case of overfitting... Do you suspect something else like a bug ?",
      "votes": null
    },
    {
      "id": "784653",
      "postDate": "03/24/2020 12:14:12",
      "content": "<p>Not exactly sure...\nI wish it was one of those 2: overfitting.. or bug...</p>\n\n<p>but here is the case/situation...</p>\n\n<p>Even other public notebooks that gave 90+ score now give a score of around 50 - 60, \nas in [steps to repro]\nfork the public notebook, \nrun them[to check that they work, and nothing is broken ], \nand submit</p>\n\n<p>what was a 90+ score, now gives 50 something! <a href=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F87729%2F44104771abbfef308011d99a7b4a6445%2Frosa.PNG?generation=1585051930680369&amp;alt=media\">pic of rosa's book forked and run</a></p>\n\n<p>this is the result I got when I forked and ran Rosa's book from here...</p>\n\n<p><a href=\"https://www.kaggle.com/wrrosa/tpu-enet-b7-densenet\">https://www.kaggle.com/wrrosa/tpu-enet-b7-densenet</a></p>",
      "rawMarkdown": "Not exactly sure...\nI wish it was one of those 2: overfitting.. or bug...\n\nbut here is the case/situation...\n\nEven other public notebooks that gave 90+ score now give a score of around 50 - 60, \nas in [steps to repro]\nfork the public notebook, \nrun them[to check that they work, and nothing is broken ], \nand submit\n\nwhat was a 90+ score, now gives 50 something! [pic of rosa's book forked and run](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F87729%2F44104771abbfef308011d99a7b4a6445%2Frosa.PNG?generation=1585051930680369&amp;alt=media)\n\nthis is the result I got when I forked and ran Rosa's book from here...\n\nhttps://www.kaggle.com/wrrosa/tpu-enet-b7-densenet",
      "votes": null
    },
    {
      "id": "785295",
      "postDate": "03/25/2020 00:29:13",
      "content": "<p><a href=\"/wrrosa\">@wrrosa</a> did you change something in your public notebook ?</p>",
      "rawMarkdown": "wrrosa did you change something in your public notebook ?",
      "votes": null
    },
    {
      "id": "785639",
      "postDate": "03/25/2020 08:20:56",
      "content": "<p>Hi Martin,\nThanks for taking the time to look at this... \nI started from scratch, and it works fine now, can we close this here...?</p>\n\n<p>As for the root cause, Not sure what it is... since I am almost out of kaggle TPU hours... I am unable to look into it this week, I am beginning to think its some mixup on my part.</p>\n\n<p>Regards,\nRony</p>",
      "rawMarkdown": "Hi Martin,\nThanks for taking the time to look at this... \nI started from scratch, and it works fine now, can we close this here...?\n\nAs for the root cause, Not sure what it is... since I am almost out of kaggle TPU hours... I am unable to look into it this week, I am beginning to think its some mixup on my part.\n\nRegards,\nRony",
      "votes": null
    },
    {
      "id": "785861",
      "postDate": "03/25/2020 13:04:16",
      "content": "<p><a href=\"/mgornergoogle\">@mgornergoogle</a> No, I didn't. <a href=\"/ronyroy\">@ronyroy</a> note that almost every submit of my notebook is scored, and therefore described at the top of nb.</p>",
      "rawMarkdown": "mgornergoogle No, I didn't. @ronyroy note that almost every submit of my notebook is scored, and therefore described at the top of nb.",
      "votes": null
    },
    {
      "id": "786118",
      "postDate": "03/25/2020 16:36:30",
      "content": "<p>OK, thanks for checking.</p>\n\n<p>To close, you can edit your top post with a not like \"Edit: issue resolved; user error\"</p>",
      "rawMarkdown": "OK, thanks for checking.\n\nTo close, you can edit your top post with a not like \"Edit: issue resolved; user error\"",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 784056,
      "author_name": "mgorner",
      "author_url": "",
      "post_date": "03/23/2020 23:00:11",
      "content": "<p>This looks like a bad case of overfitting... Do you suspect something else like a bug ?</p>",
      "votes": null,
      "replies": [
        {
          "id": 784653,
          "author_name": "ronyroy",
          "author_url": "",
          "post_date": "03/24/2020 12:14:12",
          "content": "<p>Not exactly sure...\nI wish it was one of those 2: overfitting.. or bug...</p>\n\n<p>but here is the case/situation...</p>\n\n<p>Even other public notebooks that gave 90+ score now give a score of around 50 - 60, \nas in [steps to repro]\nfork the public notebook, \nrun them[to check that they work, and nothing is broken ], \nand submit</p>\n\n<p>what was a 90+ score, now gives 50 something! <a href=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F87729%2F44104771abbfef308011d99a7b4a6445%2Frosa.PNG?generation=1585051930680369&amp;alt=media\">pic of rosa's book forked and run</a></p>\n\n<p>this is the result I got when I forked and ran Rosa's book from here...</p>\n\n<p><a href=\"https://www.kaggle.com/wrrosa/tpu-enet-b7-densenet\">https://www.kaggle.com/wrrosa/tpu-enet-b7-densenet</a></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 785295,
          "author_name": "mgorner",
          "author_url": "",
          "post_date": "03/25/2020 00:29:13",
          "content": "<p><a href=\"/wrrosa\">@wrrosa</a> did you change something in your public notebook ?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 785639,
          "author_name": "ronyroy",
          "author_url": "",
          "post_date": "03/25/2020 08:20:56",
          "content": "<p>Hi Martin,\nThanks for taking the time to look at this... \nI started from scratch, and it works fine now, can we close this here...?</p>\n\n<p>As for the root cause, Not sure what it is... since I am almost out of kaggle TPU hours... I am unable to look into it this week, I am beginning to think its some mixup on my part.</p>\n\n<p>Regards,\nRony</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 785861,
          "author_name": "wrrosa",
          "author_url": "",
          "post_date": "03/25/2020 13:04:16",
          "content": "<p><a href=\"/mgornergoogle\">@mgornergoogle</a> No, I didn't. <a href=\"/ronyroy\">@ronyroy</a> note that almost every submit of my notebook is scored, and therefore described at the top of nb.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 786118,
          "author_name": "mgorner",
          "author_url": "",
          "post_date": "03/25/2020 16:36:30",
          "content": "<p>OK, thanks for checking.</p>\n\n<p>To close, you can edit your top post with a not like \"Edit: issue resolved; user error\"</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "783231": "Just what the title says...\nAnyone else facing this..?",
    "784056": "This looks like a bad case of overfitting... Do you suspect something else like a bug ?",
    "784653": "Not exactly sure...\nI wish it was one of those 2: overfitting.. or bug...\n\nbut here is the case/situation...\n\nEven other public notebooks that gave 90+ score now give a score of around 50 - 60, \nas in [steps to repro]\nfork the public notebook, \nrun them[to check that they work, and nothing is broken ], \nand submit\n\nwhat was a 90+ score, now gives 50 something! [pic of rosa's book forked and run](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F87729%2F44104771abbfef308011d99a7b4a6445%2Frosa.PNG?generation=1585051930680369&amp;alt=media)\n\nthis is the result I got when I forked and ran Rosa's book from here...\n\nhttps://www.kaggle.com/wrrosa/tpu-enet-b7-densenet",
    "785295": "wrrosa did you change something in your public notebook ?",
    "785639": "Hi Martin,\nThanks for taking the time to look at this... \nI started from scratch, and it works fine now, can we close this here...?\n\nAs for the root cause, Not sure what it is... since I am almost out of kaggle TPU hours... I am unable to look into it this week, I am beginning to think its some mixup on my part.\n\nRegards,\nRony",
    "785861": "mgornergoogle No, I didn't. @ronyroy note that almost every submit of my notebook is scored, and therefore described at the top of nb.",
    "786118": "OK, thanks for checking.\n\nTo close, you can edit your top post with a not like \"Edit: issue resolved; user error\""
  },
  "source": "meta"
}