{
  "id": 94136,
  "title": "Fastai's val loss shows NA",
  "url": "/competitions/freesound-audio-tagging-2019/discussion/94136",
  "author_name": "",
  "post_date": "2019-06-02T13:30:15.469142200Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Same issue as in <a href=\"https://www.kaggle.com/vinayaks/2d-cnn-high-score-fast-ai\">https://www.kaggle.com/vinayaks/2d-cnn-high-score-fast-ai</a>. Does anyone have a solution? Thanks in advance.</p>\n\n<p>(Changing the docker version helps)</p>",
  "messages": [
    {
      "id": "541452",
      "postDate": "06/02/2019 13:30:15",
      "content": "<p>Same issue as in <a href=\"https://www.kaggle.com/vinayaks/2d-cnn-high-score-fast-ai\">https://www.kaggle.com/vinayaks/2d-cnn-high-score-fast-ai</a>. Does anyone have a solution? Thanks in advance.</p>\n\n<p>(Changing the docker version helps)</p>",
      "rawMarkdown": "Same issue as in https://www.kaggle.com/vinayaks/2d-cnn-high-score-fast-ai. Does anyone have a solution? Thanks in advance.\n\n(Changing the docker version helps)",
      "votes": null
    },
    {
      "id": "541527",
      "postDate": "06/02/2019 15:58:25",
      "content": "<p>On the 7th cell, the original code dose <code>.split_none()</code> which means you do not use any validation sets. You could try <code>.split_by_rand_pct(0.2)</code> or any other splitting methods for example. Hope it helps!</p>",
      "rawMarkdown": "On the 7th cell, the original code dose `.split_none()` which means you do not use any validation sets. You could try `.split_by_rand_pct(0.2)` or any other splitting methods for example. Hope it helps!",
      "votes": null
    },
    {
      "id": "541532",
      "postDate": "06/02/2019 16:08:11",
      "content": "<p>Thanks!</p>",
      "rawMarkdown": "Thanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 541527,
      "author_name": "tomomakes",
      "author_url": "",
      "post_date": "06/02/2019 15:58:25",
      "content": "<p>On the 7th cell, the original code dose <code>.split_none()</code> which means you do not use any validation sets. You could try <code>.split_by_rand_pct(0.2)</code> or any other splitting methods for example. Hope it helps!</p>",
      "votes": null,
      "replies": [
        {
          "id": 541532,
          "author_name": "shujian",
          "author_url": "",
          "post_date": "06/02/2019 16:08:11",
          "content": "<p>Thanks!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "541452": "Same issue as in https://www.kaggle.com/vinayaks/2d-cnn-high-score-fast-ai. Does anyone have a solution? Thanks in advance.\n\n(Changing the docker version helps)",
    "541527": "On the 7th cell, the original code dose `.split_none()` which means you do not use any validation sets. You could try `.split_by_rand_pct(0.2)` or any other splitting methods for example. Hope it helps!",
    "541532": "Thanks!"
  },
  "source": "meta"
}