{
  "id": 95408,
  "title": "[Lesson learnt]Remember to spend time to validate your kernel for stage-2",
  "url": "/competitions/freesound-audio-tagging-2019/discussion/95408",
  "author_name": "",
  "post_date": "2019-06-12T04:17:10.860614Z",
  "votes": 7,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi all</p>\n\n<h2>First, Hope all teams have good result for this competition</h2>\n\n<p>I want to share what I learnt in this competition.\nI joined this competition lately, about one week before the end of competition, \nFor a quickly start, I used <a href=\"/daisukelab\">@daisukelab</a> 's pre-process data to build model and training (Many thanks for good pre-process data sharing)\nIn this week, I focus on LB score improvement (Try different model, ensemble, mix-up, K-fold, Loss function change...), until last hour to competition end, I suddenly realized that I didn't pre-process \"test\" data in my kernel (I forgot... I keep using pre-process dataset...), I only have last submit chance ... fortunately, I submit my last kernel in time (but score is not my best score)...\nThis is not my first kernel competition, but I still make this kind of mistake.</p>\n\n<h2>Remember to spend time to validate your kernel for stage - 2</h2>\n\n<ul>\n<li>Test data must be re-process in kernel </li>\n<li>Kernel runtime issue, do what you can do to reduce the run-time</li>\n<li>Kernel memory, always have free kernel memory for stage-2 dataset (In previous competition, my stage-2 kernel all crash, because of memory...)</li>\n</ul>",
  "messages": [
    {
      "id": "550820",
      "postDate": "06/12/2019 04:17:10",
      "content": "<p>Hi all</p>\n\n<h2>First, Hope all teams have good result for this competition</h2>\n\n<p>I want to share what I learnt in this competition.\nI joined this competition lately, about one week before the end of competition, \nFor a quickly start, I used <a href=\"/daisukelab\">@daisukelab</a> 's pre-process data to build model and training (Many thanks for good pre-process data sharing)\nIn this week, I focus on LB score improvement (Try different model, ensemble, mix-up, K-fold, Loss function change...), until last hour to competition end, I suddenly realized that I didn't pre-process \"test\" data in my kernel (I forgot... I keep using pre-process dataset...), I only have last submit chance ... fortunately, I submit my last kernel in time (but score is not my best score)...\nThis is not my first kernel competition, but I still make this kind of mistake.</p>\n\n<h2>Remember to spend time to validate your kernel for stage - 2</h2>\n\n<ul>\n<li>Test data must be re-process in kernel </li>\n<li>Kernel runtime issue, do what you can do to reduce the run-time</li>\n<li>Kernel memory, always have free kernel memory for stage-2 dataset (In previous competition, my stage-2 kernel all crash, because of memory...)</li>\n</ul>",
      "rawMarkdown": "Hi all\n##First, Hope all teams have good result for this competition\n\nI want to share what I learnt in this competition.\nI joined this competition lately, about one week before the end of competition, \nFor a quickly start, I used @daisukelab 's pre-process data to build model and training (Many thanks for good pre-process data sharing)\nIn this week, I focus on LB score improvement (Try different model, ensemble, mix-up, K-fold, Loss function change...), until last hour to competition end, I suddenly realized that I didn't pre-process \"test\" data in my kernel (I forgot... I keep using pre-process dataset...), I only have last submit chance ... fortunately, I submit my last kernel in time (but score is not my best score)...\nThis is not my first kernel competition, but I still make this kind of mistake.\n\n##Remember to spend time to validate your kernel for stage - 2\n- Test data must be re-process in kernel \n- Kernel runtime issue, do what you can do to reduce the run-time\n- Kernel memory, always have free kernel memory for stage-2 dataset (In previous competition, my stage-2 kernel all crash, because of memory...)",
      "votes": null
    },
    {
      "id": "550828",
      "postDate": "06/12/2019 04:24:55",
      "content": "<p>Thanks for sharing!</p>",
      "rawMarkdown": "Thanks for sharing!",
      "votes": null
    },
    {
      "id": "550847",
      "postDate": "06/12/2019 04:46:12",
      "content": "<p>I remembered the same just couple of hours before the competition end and was able to add stage-2 test pre-process in time. For now both versions are running fine...not sure if any other kernel related issues will crop up</p>",
      "rawMarkdown": "I remembered the same just couple of hours before the competition end and was able to add stage-2 test pre-process in time. For now both versions are running fine...not sure if any other kernel related issues will crop up",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 550828,
      "author_name": "tayorm",
      "author_url": "",
      "post_date": "06/12/2019 04:24:55",
      "content": "<p>Thanks for sharing!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 550847,
      "author_name": "sandy1112",
      "author_url": "",
      "post_date": "06/12/2019 04:46:12",
      "content": "<p>I remembered the same just couple of hours before the competition end and was able to add stage-2 test pre-process in time. For now both versions are running fine...not sure if any other kernel related issues will crop up</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "550820": "Hi all\n##First, Hope all teams have good result for this competition\n\nI want to share what I learnt in this competition.\nI joined this competition lately, about one week before the end of competition, \nFor a quickly start, I used @daisukelab 's pre-process data to build model and training (Many thanks for good pre-process data sharing)\nIn this week, I focus on LB score improvement (Try different model, ensemble, mix-up, K-fold, Loss function change...), until last hour to competition end, I suddenly realized that I didn't pre-process \"test\" data in my kernel (I forgot... I keep using pre-process dataset...), I only have last submit chance ... fortunately, I submit my last kernel in time (but score is not my best score)...\nThis is not my first kernel competition, but I still make this kind of mistake.\n\n##Remember to spend time to validate your kernel for stage - 2\n- Test data must be re-process in kernel \n- Kernel runtime issue, do what you can do to reduce the run-time\n- Kernel memory, always have free kernel memory for stage-2 dataset (In previous competition, my stage-2 kernel all crash, because of memory...)",
    "550828": "Thanks for sharing!",
    "550847": "I remembered the same just couple of hours before the competition end and was able to add stage-2 test pre-process in time. For now both versions are running fine...not sure if any other kernel related issues will crop up"
  },
  "source": "meta"
}