{
  "id": 145209,
  "title": "Dear competition organizers: just one test image please?",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/145209",
  "author_name": "ilovescience",
  "post_date": "2020-04-22T09:43:10.975000",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>It is hard to create an inference kernel when we don't have access to a <code>test_images</code> folder in our directory with no example test images. Is it possible to have even one? I would be satisfied with three images. At least, then we can make sure our inference kernels run properly.</p>\n\n<p>In these synchronous code competitions, it is already tough to make sure our inference kernels work on the private test set when there are no useful error messages when it fails. Now, this competition takes it to a whole new level. We really just have to guess and hope our inference kernels work, and I expect it will take longer than usual to get a working inference kernel.</p>\n\n<p>I hope the competition organizers strongly consider adding even just one test image to the kernel environment to check if our kernels work.</p>",
  "messages": [
    {
      "id": 816343,
      "postDate": "2020-04-22T09:43:10.977Z",
      "content": "<p>It is hard to create an inference kernel when we don't have access to a <code>test_images</code> folder in our directory with no example test images. Is it possible to have even one? I would be satisfied with three images. At least, then we can make sure our inference kernels run properly.</p>\n\n<p>In these synchronous code competitions, it is already tough to make sure our inference kernels work on the private test set when there are no useful error messages when it fails. Now, this competition takes it to a whole new level. We really just have to guess and hope our inference kernels work, and I expect it will take longer than usual to get a working inference kernel.</p>\n\n<p>I hope the competition organizers strongly consider adding even just one test image to the kernel environment to check if our kernels work.</p>",
      "rawMarkdown": "It is hard to create an inference kernel when we don't have access to a `test_images` folder in our directory with no example test images. Is it possible to have even one? I would be satisfied with three images. At least, then we can make sure our inference kernels run properly.\n\nIn these synchronous code competitions, it is already tough to make sure our inference kernels work on the private test set when there are no useful error messages when it fails. Now, this competition takes it to a whole new level. We really just have to guess and hope our inference kernels work, and I expect it will take longer than usual to get a working inference kernel.\n\nI hope the competition organizers strongly consider adding even just one test image to the kernel environment to check if our kernels work.",
      "votes": 3
    },
    {
      "id": 817705,
      "postDate": "2020-04-23T10:54:41.983Z",
      "content": "<p>It would be much more convenient to run code with 1-2 test examples than to write \"if else\" blocks</p>",
      "rawMarkdown": "It would be much more convenient to run code with 1-2 test examples than to write \"if else\" blocks"
    },
    {
      "id": 817646,
      "postDate": "2020-04-23T09:36:45.540Z",
      "content": "<p>Just do something like this <a href=\"https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/145219\">https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/145219</a></p>\n\n<p>If you want a validation set, extract it from the training data</p>",
      "rawMarkdown": "Just do something like this https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/145219\n\nIf you want a validation set, extract it from the training data"
    },
    {
      "id": 816384,
      "postDate": "2020-04-22T10:38:11.070Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 817705,
      "author_name": "Artur Fattakhov (MIPT DIHT)",
      "author_url": "",
      "post_date": "2020-04-23T10:54:41.983000",
      "content": "<p>It would be much more convenient to run code with 1-2 test examples than to write \"if else\" blocks</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 817646,
      "author_name": "Matt",
      "author_url": "",
      "post_date": "2020-04-23T09:36:45.540000",
      "content": "<p>Just do something like this <a href=\"https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/145219\">https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/145219</a></p>\n\n<p>If you want a validation set, extract it from the training data</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 816384,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-04-22T10:38:11.070000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "816343": "It is hard to create an inference kernel when we don't have access to a `test_images` folder in our directory with no example test images. Is it possible to have even one? I would be satisfied with three images. At least, then we can make sure our inference kernels run properly.\n\nIn these synchronous code competitions, it is already tough to make sure our inference kernels work on the private test set when there are no useful error messages when it fails. Now, this competition takes it to a whole new level. We really just have to guess and hope our inference kernels work, and I expect it will take longer than usual to get a working inference kernel.\n\nI hope the competition organizers strongly consider adding even just one test image to the kernel environment to check if our kernels work.",
    "817705": "It would be much more convenient to run code with 1-2 test examples than to write \"if else\" blocks",
    "817646": "Just do something like this https://www.kaggle.com/c/prostate-cancer-grade-assessment/discussion/145219\n\nIf you want a validation set, extract it from the training data",
    "816384": ""
  }
}