{
  "id": 493895,
  "title": " Empty test_soundscapes - Submission issues",
  "url": "/competitions/birdclef-2024/discussion/493895",
  "author_name": "",
  "post_date": "2024-04-15T09:03:03.074039500Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Good morning everyone,</p>\n<p>it is the first time I participate to a bird clef competition. I feel like the competition setup and relative instructions are not that clear. Maybe it is because I was not able to fetch the suitable information.</p>\n<p>For the moment, I wasted a lot of time trying to guess what I should do to submit something correctly.</p>\n<p>Up to now, in my submission logs, I do not manage to see my test_soundscapes folder populated by any file apart from its readme.txt.<br>\n BTW, in this file I can read ' Only the hidden rerun copy of the test soundscape dirctory will only be populated.' which sounds a bit cryptic to me. I also read somewhere that the usage of no accelerator entails a time limitation of one minute for the kernel.</p>\n<p>Could we please have a text with (or link to) clear guidance for the required properties of the notebooks to be submitted to this competition?<br>\nfor instance</p>\n<ul>\n<li>what accelerator options may be used ?</li>\n<li>is the sample_submission.csv updated when we submit ?</li>\n<li>what is the notion 'the hidden rerun copy' ?</li>\n<li>can we expect to get any relevant feedback from our submitted notebooks' logs ? </li>\n</ul>\n<p>I can say no such information is in  the pinned discussions.</p>\n<p>In science, reproducibility is an important notion and I think well defined concepts help a lot in this direction.<br>\nI feel very bad when I end up mimicking other's stuff without knowing what I can modify safely.</p>\n<p>However, I understand the organizers have a lot of work and I am not blaming them personally. I am just explaining some frustration and pointing some directions in which one could go to avoid useless work.</p>\n<p>Best,</p>",
  "messages": [
    {
      "id": "2753044",
      "postDate": "04/15/2024 09:03:03",
      "content": "<p>Good morning everyone,</p>\n<p>it is the first time I participate to a bird clef competition. I feel like the competition setup and relative instructions are not that clear. Maybe it is because I was not able to fetch the suitable information.</p>\n<p>For the moment, I wasted a lot of time trying to guess what I should do to submit something correctly.</p>\n<p>Up to now, in my submission logs, I do not manage to see my test_soundscapes folder populated by any file apart from its readme.txt.<br>\n BTW, in this file I can read ' Only the hidden rerun copy of the test soundscape dirctory will only be populated.' which sounds a bit cryptic to me. I also read somewhere that the usage of no accelerator entails a time limitation of one minute for the kernel.</p>\n<p>Could we please have a text with (or link to) clear guidance for the required properties of the notebooks to be submitted to this competition?<br>\nfor instance</p>\n<ul>\n<li>what accelerator options may be used ?</li>\n<li>is the sample_submission.csv updated when we submit ?</li>\n<li>what is the notion 'the hidden rerun copy' ?</li>\n<li>can we expect to get any relevant feedback from our submitted notebooks' logs ? </li>\n</ul>\n<p>I can say no such information is in  the pinned discussions.</p>\n<p>In science, reproducibility is an important notion and I think well defined concepts help a lot in this direction.<br>\nI feel very bad when I end up mimicking other's stuff without knowing what I can modify safely.</p>\n<p>However, I understand the organizers have a lot of work and I am not blaming them personally. I am just explaining some frustration and pointing some directions in which one could go to avoid useless work.</p>\n<p>Best,</p>",
      "rawMarkdown": "Good morning everyone,\n\nit is the first time I participate to a bird clef competition. I feel like the competition setup and relative instructions are not that clear. Maybe it is because I was not able to fetch the suitable information.\n\nFor the moment, I wasted a lot of time trying to guess what I should do to submit something correctly.\n\nUp to now, in my submission logs, I do not manage to see my test_soundscapes folder populated by any file apart from its readme.txt.\n BTW, in this file I can read ' Only the hidden rerun copy of the test soundscape dirctory will only be populated.' which sounds a bit cryptic to me. I also read somewhere that the usage of no accelerator entails a time limitation of one minute for the kernel.\n\nCould we please have a text with (or link to) clear guidance for the required properties of the notebooks to be submitted to this competition?\nfor instance\n- what accelerator options may be used ?\n- is the sample_submission.csv updated when we submit ?\n- what is the notion 'the hidden rerun copy' ?\n- can we expect to get any relevant feedback from our submitted notebooks' logs ? \n\nI can say no such information is in  the pinned discussions.\n\nIn science, reproducibility is an important notion and I think well defined concepts help a lot in this direction.\nI feel very bad when I end up mimicking other's stuff without knowing what I can modify safely.\n\nHowever, I understand the organizers have a lot of work and I am not blaming them personally. I am just explaining some frustration and pointing some directions in which one could go to avoid useless work.\n\n\nBest,",
      "votes": null
    },
    {
      "id": "2754002",
      "postDate": "04/15/2024 18:44:02",
      "content": "<p>Hello, and welcome to BirdCLEF! </p>\n<p>This is a 'hidden test set' competition; there is limited feedback from the test runs to limit overfit to the test data. So pretty much all you get is leaderboard score and failure indicators.</p>\n<p>The submission should run on CPU, as discussed in the 'Code Requirements' section of the competition overview.</p>\n<p>We strongly suggest starting from a working submission example, such as this one, instead of coding the submission logic from scratch. It's (hopefully) the least interesting part of the problem!<br>\n<a href=\"https://www.kaggle.com/code/myso1987/how-to-submit-to-birdclef-2024\" target=\"_blank\">https://www.kaggle.com/code/myso1987/how-to-submit-to-birdclef-2024</a></p>",
      "rawMarkdown": "Hello, and welcome to BirdCLEF! \n\nThis is a 'hidden test set' competition; there is limited feedback from the test runs to limit overfit to the test data. So pretty much all you get is leaderboard score and failure indicators.\n\nThe submission should run on CPU, as discussed in the 'Code Requirements' section of the competition overview.\n\nWe strongly suggest starting from a working submission example, such as this one, instead of coding the submission logic from scratch. It's (hopefully) the least interesting part of the problem!\nhttps://www.kaggle.com/code/myso1987/how-to-submit-to-birdclef-2024",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2754002,
      "author_name": "tomdenton",
      "author_url": "",
      "post_date": "04/15/2024 18:44:02",
      "content": "<p>Hello, and welcome to BirdCLEF! </p>\n<p>This is a 'hidden test set' competition; there is limited feedback from the test runs to limit overfit to the test data. So pretty much all you get is leaderboard score and failure indicators.</p>\n<p>The submission should run on CPU, as discussed in the 'Code Requirements' section of the competition overview.</p>\n<p>We strongly suggest starting from a working submission example, such as this one, instead of coding the submission logic from scratch. It's (hopefully) the least interesting part of the problem!<br>\n<a href=\"https://www.kaggle.com/code/myso1987/how-to-submit-to-birdclef-2024\" target=\"_blank\">https://www.kaggle.com/code/myso1987/how-to-submit-to-birdclef-2024</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2753044": "Good morning everyone,\n\nit is the first time I participate to a bird clef competition. I feel like the competition setup and relative instructions are not that clear. Maybe it is because I was not able to fetch the suitable information.\n\nFor the moment, I wasted a lot of time trying to guess what I should do to submit something correctly.\n\nUp to now, in my submission logs, I do not manage to see my test_soundscapes folder populated by any file apart from its readme.txt.\n BTW, in this file I can read ' Only the hidden rerun copy of the test soundscape dirctory will only be populated.' which sounds a bit cryptic to me. I also read somewhere that the usage of no accelerator entails a time limitation of one minute for the kernel.\n\nCould we please have a text with (or link to) clear guidance for the required properties of the notebooks to be submitted to this competition?\nfor instance\n- what accelerator options may be used ?\n- is the sample_submission.csv updated when we submit ?\n- what is the notion 'the hidden rerun copy' ?\n- can we expect to get any relevant feedback from our submitted notebooks' logs ? \n\nI can say no such information is in  the pinned discussions.\n\nIn science, reproducibility is an important notion and I think well defined concepts help a lot in this direction.\nI feel very bad when I end up mimicking other's stuff without knowing what I can modify safely.\n\nHowever, I understand the organizers have a lot of work and I am not blaming them personally. I am just explaining some frustration and pointing some directions in which one could go to avoid useless work.\n\n\nBest,",
    "2754002": "Hello, and welcome to BirdCLEF! \n\nThis is a 'hidden test set' competition; there is limited feedback from the test runs to limit overfit to the test data. So pretty much all you get is leaderboard score and failure indicators.\n\nThe submission should run on CPU, as discussed in the 'Code Requirements' section of the competition overview.\n\nWe strongly suggest starting from a working submission example, such as this one, instead of coding the submission logic from scratch. It's (hopefully) the least interesting part of the problem!\nhttps://www.kaggle.com/code/myso1987/how-to-submit-to-birdclef-2024"
  },
  "source": "meta"
}