{
  "id": 88597,
  "title": "Can Keras+Public Kernel reach 0.600?",
  "url": "/competitions/imet-2019-fgvc6/discussion/88597",
  "author_name": "",
  "post_date": "2019-04-09T15:54:45.912877100Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Is there anyone who is able to reach 0.600++ using Keras within 9 hours on public kernel??</p>",
  "messages": [
    {
      "id": "510927",
      "postDate": "04/09/2019 15:54:45",
      "content": "<p>Is there anyone who is able to reach 0.600++ using Keras within 9 hours on public kernel??</p>",
      "rawMarkdown": "Is there anyone who is able to reach 0.600++ using Keras within 9 hours on public kernel??",
      "votes": null
    },
    {
      "id": "516380",
      "postDate": "04/14/2019 04:24:07",
      "content": "<p>How do you know that the people at the top use pytorch?</p>",
      "rawMarkdown": "How do you know that the people at the top use pytorch?",
      "votes": null
    },
    {
      "id": "517447",
      "postDate": "04/16/2019 03:28:52",
      "content": "<p>Haha you are right. I just made a guess. I now change the content to not make confusion.</p>\n\n<p>Thanks!</p>",
      "rawMarkdown": "Haha you are right. I just made a guess. I now change the content to not make confusion.\n\nThanks!",
      "votes": null
    },
    {
      "id": "518355",
      "postDate": "04/17/2019 05:35:24",
      "content": "<p><a href=\"https://medium.com/comet-ml/approach-pre-trained-deep-learning-models-with-caution-9f0ff739010c\">Approach pre-trained deep learning models with caution</a> Hi, I think this article could be helpful, for it claimed that &gt; Keras models using batch normalization can be unreliable. For some models, forward-pass evaluations (with gradients supposedly off) still result in weights changing at inference time.&gt; </p>",
      "rawMarkdown": "[Approach pre-trained deep learning models with caution](https://medium.com/comet-ml/approach-pre-trained-deep-learning-models-with-caution-9f0ff739010c) Hi, I think this article could be helpful, for it claimed that &gt; Keras models using batch normalization can be unreliable. For some models, forward-pass evaluations (with gradients supposedly off) still result in weights changing at inference time.&gt;",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 516380,
      "author_name": "",
      "author_url": "",
      "post_date": "04/14/2019 04:24:07",
      "content": "<p>How do you know that the people at the top use pytorch?</p>",
      "votes": null,
      "replies": [
        {
          "id": 517447,
          "author_name": "ratthachat",
          "author_url": "",
          "post_date": "04/16/2019 03:28:52",
          "content": "<p>Haha you are right. I just made a guess. I now change the content to not make confusion.</p>\n\n<p>Thanks!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 518355,
      "author_name": "leonshangguan",
      "author_url": "",
      "post_date": "04/17/2019 05:35:24",
      "content": "<p><a href=\"https://medium.com/comet-ml/approach-pre-trained-deep-learning-models-with-caution-9f0ff739010c\">Approach pre-trained deep learning models with caution</a> Hi, I think this article could be helpful, for it claimed that &gt; Keras models using batch normalization can be unreliable. For some models, forward-pass evaluations (with gradients supposedly off) still result in weights changing at inference time.&gt; </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "510927": "Is there anyone who is able to reach 0.600++ using Keras within 9 hours on public kernel??",
    "516380": "How do you know that the people at the top use pytorch?",
    "517447": "Haha you are right. I just made a guess. I now change the content to not make confusion.\n\nThanks!",
    "518355": "[Approach pre-trained deep learning models with caution](https://medium.com/comet-ml/approach-pre-trained-deep-learning-models-with-caution-9f0ff739010c) Hi, I think this article could be helpful, for it claimed that &gt; Keras models using batch normalization can be unreliable. For some models, forward-pass evaluations (with gradients supposedly off) still result in weights changing at inference time.&gt;"
  },
  "source": "meta"
}