{
  "id": 95924,
  "title": "1st place solution released on a Github ",
  "url": "/competitions/freesound-audio-tagging-2019/discussion/95924",
  "author_name": "Ruslan Baikulov",
  "post_date": "2019-06-16T11:49:33.094000",
  "votes": 66,
  "comment_count": 12,
  "views": 0,
  "content": "<p>Hello everyone, thanks for the competition.</p>\n\n<p>Link to solution <a href=\"https://github.com/lRomul/argus-freesound\">https://github.com/lRomul/argus-freesound</a></p>\n\n<p>Key points:</p>\n\n<ul>\n<li>CNN model with attention, skip connections and auxiliary classifiers </li>\n<li>SpecAugment, Mixup augmentations </li>\n<li>Ensemble with MLP second-level model and geometric mean blending </li>\n</ul>\n\n<p>More details you can find in the repository <a href=\"https://github.com/lRomul/argus-freesound/blob/master/README.md\">README</a>.</p>",
  "messages": [
    {
      "id": 553818,
      "postDate": "2019-06-16T11:49:33.093Z",
      "content": "<p>Hello everyone, thanks for the competition.</p>\n\n<p>Link to solution <a href=\"https://github.com/lRomul/argus-freesound\">https://github.com/lRomul/argus-freesound</a></p>\n\n<p>Key points:</p>\n\n<ul>\n<li>CNN model with attention, skip connections and auxiliary classifiers </li>\n<li>SpecAugment, Mixup augmentations </li>\n<li>Ensemble with MLP second-level model and geometric mean blending </li>\n</ul>\n\n<p>More details you can find in the repository <a href=\"https://github.com/lRomul/argus-freesound/blob/master/README.md\">README</a>.</p>",
      "rawMarkdown": "Hello everyone, thanks for the competition.\n\nLink to solution https://github.com/lRomul/argus-freesound\n\nKey points:\n\n* CNN model with attention, skip connections and auxiliary classifiers \n* SpecAugment, Mixup augmentations \n* Ensemble with MLP second-level model and geometric mean blending \n\nMore details you can find in the repository [README](https://github.com/lRomul/argus-freesound/blob/master/README.md).",
      "votes": 66
    },
    {
      "id": 564229,
      "postDate": "2019-06-29T06:51:42.023Z",
      "content": "<p>Congratulations <a href=\"/romul0212\">@romul0212</a> !\nAnd thanks for sharing your solution.</p>",
      "rawMarkdown": "Congratulations @romul0212 !\nAnd thanks for sharing your solution.",
      "votes": 1
    },
    {
      "id": 554805,
      "postDate": "2019-06-18T03:23:01.817Z",
      "content": "<p>Thanks for sharing not only your codes but also lab_journal, it's quite interesting!</p>",
      "rawMarkdown": "Thanks for sharing not only your codes but also lab_journal, it's quite interesting!",
      "votes": 1
    },
    {
      "id": 554746,
      "postDate": "2019-06-18T01:18:46.583Z",
      "content": "<p>Congratulations！</p>",
      "rawMarkdown": "Congratulations！",
      "votes": 1
    },
    {
      "id": 554743,
      "postDate": "2019-06-18T01:16:12.177Z",
      "content": "<p>Congratulations.\nSpecAugment didn't work for us.</p>",
      "rawMarkdown": "Congratulations.\nSpecAugment didn't work for us.",
      "votes": 1
    },
    {
      "id": 554722,
      "postDate": "2019-06-18T00:24:01.210Z",
      "content": "<p>Congratulations (you jumped quite high in private LB before it was withdrawn), and thanks for sharing a well written and detailed report. I have a question on your <a href=\"https://docs.google.com/spreadsheets/d/1uOp2Du3CROtpg7TuSFmSejyXQe2Dp8DGh5Dm5onBWfc/edit?usp=sharing\">laboratory journal</a>. Does it represent all of your last lvl1 models included in the final submission. What is the range of CV and public LB (if applicable) for them?</p>",
      "rawMarkdown": "Congratulations (you jumped quite high in private LB before it was withdrawn), and thanks for sharing a well written and detailed report. I have a question on your [laboratory journal](https://docs.google.com/spreadsheets/d/1uOp2Du3CROtpg7TuSFmSejyXQe2Dp8DGh5Dm5onBWfc/edit?usp=sharing). Does it represent all of your last lvl1 models included in the final submission. What is the range of CV and public LB (if applicable) for them?",
      "votes": 1,
      "replies": [
        {
          "id": 554879,
          "postDate": "2019-06-18T06:19:53.853Z",
          "content": "<p>lvl1 models in the final submission are in the sheet \"Train 3\". Information about ensembling submissions is in the sheet \"Stacking\".</p>",
          "rawMarkdown": "lvl1 models in the final submission are in the sheet \"Train 3\". Information about ensembling submissions is in the sheet \"Stacking\".",
          "votes": 1
        }
      ]
    },
    {
      "id": 1094204,
      "postDate": "2020-11-28T12:06:08.560Z",
      "content": "<p>Congratulation <a href=\"https://www.kaggle.com/romul0212\" target=\"_blank\">@romul0212</a>! And thank you for sharing your excellent work with the well informative repo 👍!</p>",
      "rawMarkdown": "Congratulation @romul0212! And thank you for sharing your excellent work with the well informative repo 👍!"
    },
    {
      "id": 742829,
      "postDate": "2020-02-11T15:09:25.503Z",
      "content": "<p>hi, I just wanted to know whether one must need a 2080 ti to execute the 1st place solution or the graphics memory should be larger than 12GB is a necessity?</p>",
      "rawMarkdown": "hi, I just wanted to know whether one must need a 2080 ti to execute the 1st place solution or the graphics memory should be larger than 12GB is a necessity?"
    },
    {
      "id": 554482,
      "postDate": "2019-06-17T13:27:48.980Z",
      "content": "<p>Is it allowed by rules to relabel the data?</p>",
      "rawMarkdown": "Is it allowed by rules to relabel the data?",
      "replies": [
        {
          "id": 554547,
          "postDate": "2019-06-17T16:19:20.627Z",
          "content": "<p>Yes, but only train data</p>",
          "rawMarkdown": "Yes, but only train data"
        }
      ]
    },
    {
      "id": 555057,
      "postDate": "2019-06-18T11:33:27.657Z",
      "rawMarkdown": "",
      "isDeleted": true
    },
    {
      "id": 586969,
      "postDate": "2019-07-30T00:39:44.213Z",
      "content": "<p>Thanks, and congrats!</p>",
      "rawMarkdown": "Thanks, and congrats!"
    }
  ],
  "comments": [
    {
      "id": 564229,
      "author_name": "Vopani",
      "author_url": "",
      "post_date": "2019-06-29T06:51:42.023000",
      "content": "<p>Congratulations <a href=\"/romul0212\">@romul0212</a> !\nAnd thanks for sharing your solution.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 554805,
      "author_name": "daisukelab",
      "author_url": "",
      "post_date": "2019-06-18T03:23:01.817000",
      "content": "<p>Thanks for sharing not only your codes but also lab_journal, it's quite interesting!</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 554746,
      "author_name": "qrfaction",
      "author_url": "",
      "post_date": "2019-06-18T01:18:46.583000",
      "content": "<p>Congratulations！</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 554743,
      "author_name": "Gary",
      "author_url": "",
      "post_date": "2019-06-18T01:16:12.177000",
      "content": "<p>Congratulations.\nSpecAugment didn't work for us.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 554722,
      "author_name": "Iafoss",
      "author_url": "",
      "post_date": "2019-06-18T00:24:01.210000",
      "content": "<p>Congratulations (you jumped quite high in private LB before it was withdrawn), and thanks for sharing a well written and detailed report. I have a question on your <a href=\"https://docs.google.com/spreadsheets/d/1uOp2Du3CROtpg7TuSFmSejyXQe2Dp8DGh5Dm5onBWfc/edit?usp=sharing\">laboratory journal</a>. Does it represent all of your last lvl1 models included in the final submission. What is the range of CV and public LB (if applicable) for them?</p>",
      "votes": 1,
      "replies": [
        {
          "id": 554879,
          "author_name": "Ruslan Baikulov",
          "author_url": "",
          "post_date": "2019-06-18T06:19:53.853000",
          "content": "<p>lvl1 models in the final submission are in the sheet \"Train 3\". Information about ensembling submissions is in the sheet \"Stacking\".</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1094204,
      "author_name": "Beans",
      "author_url": "",
      "post_date": "2020-11-28T12:06:08.560000",
      "content": "<p>Congratulation <a href=\"https://www.kaggle.com/romul0212\" target=\"_blank\">@romul0212</a>! And thank you for sharing your excellent work with the well informative repo 👍!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 742829,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-02-11T15:09:25.503000",
      "content": "<p>hi, I just wanted to know whether one must need a 2080 ti to execute the 1st place solution or the graphics memory should be larger than 12GB is a necessity?</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 554482,
      "author_name": "Jing Liu",
      "author_url": "",
      "post_date": "2019-06-17T13:27:48.980000",
      "content": "<p>Is it allowed by rules to relabel the data?</p>",
      "votes": 0,
      "replies": [
        {
          "id": 554547,
          "author_name": "Ruslan Baikulov",
          "author_url": "",
          "post_date": "2019-06-17T16:19:20.627000",
          "content": "<p>Yes, but only train data</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 555057,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-06-18T11:33:27.657000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 586969,
      "author_name": "Keagan",
      "author_url": "",
      "post_date": "2019-07-30T00:39:44.213000",
      "content": "<p>Thanks, and congrats!</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "553818": "Hello everyone, thanks for the competition.\n\nLink to solution https://github.com/lRomul/argus-freesound\n\nKey points:\n\n* CNN model with attention, skip connections and auxiliary classifiers \n* SpecAugment, Mixup augmentations \n* Ensemble with MLP second-level model and geometric mean blending \n\nMore details you can find in the repository [README](https://github.com/lRomul/argus-freesound/blob/master/README.md).",
    "564229": "Congratulations @romul0212 !\nAnd thanks for sharing your solution.",
    "554805": "Thanks for sharing not only your codes but also lab_journal, it's quite interesting!",
    "554746": "Congratulations！",
    "554743": "Congratulations.\nSpecAugment didn't work for us.",
    "554722": "Congratulations (you jumped quite high in private LB before it was withdrawn), and thanks for sharing a well written and detailed report. I have a question on your [laboratory journal](https://docs.google.com/spreadsheets/d/1uOp2Du3CROtpg7TuSFmSejyXQe2Dp8DGh5Dm5onBWfc/edit?usp=sharing). Does it represent all of your last lvl1 models included in the final submission. What is the range of CV and public LB (if applicable) for them?",
    "1094204": "Congratulation @romul0212! And thank you for sharing your excellent work with the well informative repo 👍!",
    "742829": "hi, I just wanted to know whether one must need a 2080 ti to execute the 1st place solution or the graphics memory should be larger than 12GB is a necessity?",
    "554482": "Is it allowed by rules to relabel the data?",
    "555057": "",
    "586969": "Thanks, and congrats!"
  }
}