{
  "id": 407220,
  "title": "Did you manage to get local evaluation to work?",
  "url": "/competitions/cafa-5-protein-function-prediction/discussion/407220",
  "author_name": "",
  "post_date": "2023-05-05T14:51:43.654398200Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I'm trying to run local evaluation of the train terms against themselves (i.e. get the optimal score), and it is taking forever. Just running the propagation of maxes from parents to children (did I understand right?) function on all files takes twenty minutes (and I'm running a beefy cloud machine), and I have yet to see the last evaluation stage finish.</p>\n<p>Has anyone gotten to run evaluation locally in a reasonable time?<br>\nIs there way to split my predictions file so that the evaluation will be faster? (this hinted by the evaluation api taking a predictions folder, and not just a file)</p>\n<p>[EDIT:] I got results back, it took 93 minutes :(</p>",
  "messages": [
    {
      "id": "2246929",
      "postDate": "05/05/2023 14:51:43",
      "content": "<p>I'm trying to run local evaluation of the train terms against themselves (i.e. get the optimal score), and it is taking forever. Just running the propagation of maxes from parents to children (did I understand right?) function on all files takes twenty minutes (and I'm running a beefy cloud machine), and I have yet to see the last evaluation stage finish.</p>\n<p>Has anyone gotten to run evaluation locally in a reasonable time?<br>\nIs there way to split my predictions file so that the evaluation will be faster? (this hinted by the evaluation api taking a predictions folder, and not just a file)</p>\n<p>[EDIT:] I got results back, it took 93 minutes :(</p>",
      "rawMarkdown": "I'm trying to run local evaluation of the train terms against themselves (i.e. get the optimal score), and it is taking forever. Just running the propagation of maxes from parents to children (did I understand right?) function on all files takes twenty minutes (and I'm running a beefy cloud machine), and I have yet to see the last evaluation stage finish.\n\nHas anyone gotten to run evaluation locally in a reasonable time?\nIs there way to split my predictions file so that the evaluation will be faster? (this hinted by the evaluation api taking a predictions folder, and not just a file)\n\n[EDIT:] I got results back, it took 93 minutes :(",
      "votes": null
    },
    {
      "id": "2247130",
      "postDate": "05/05/2023 18:26:04",
      "content": "<p>Same here. I'm resorting to using a small percentage of proteins/terms.</p>",
      "rawMarkdown": "Same here. I'm resorting to using a small percentage of proteins/terms.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2247130,
      "author_name": "mtinti",
      "author_url": "",
      "post_date": "05/05/2023 18:26:04",
      "content": "<p>Same here. I'm resorting to using a small percentage of proteins/terms.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2246929": "I'm trying to run local evaluation of the train terms against themselves (i.e. get the optimal score), and it is taking forever. Just running the propagation of maxes from parents to children (did I understand right?) function on all files takes twenty minutes (and I'm running a beefy cloud machine), and I have yet to see the last evaluation stage finish.\n\nHas anyone gotten to run evaluation locally in a reasonable time?\nIs there way to split my predictions file so that the evaluation will be faster? (this hinted by the evaluation api taking a predictions folder, and not just a file)\n\n[EDIT:] I got results back, it took 93 minutes :(",
    "2247130": "Same here. I'm resorting to using a small percentage of proteins/terms."
  },
  "source": "meta"
}