{
  "id": 492282,
  "title": "Inference time for a single model?",
  "url": "/competitions/birdclef-2024/discussion/492282",
  "author_name": "",
  "post_date": "2024-04-09T05:46:20.802282500Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Given that this competition only allows cpu submissions which are restricted to 2 hours, its very important to build proper pipelines and made make optimize inference as much as possible, my single rough efficientnetb0 took around 1 hour which needs a lot of optimization, I would like to know how efficient others submissions are and number of models we might be able to fit within the 2 hour submission window.</p>",
  "messages": [
    {
      "id": "2742853",
      "postDate": "04/09/2024 05:46:20",
      "content": "<p>Given that this competition only allows cpu submissions which are restricted to 2 hours, its very important to build proper pipelines and made make optimize inference as much as possible, my single rough efficientnetb0 took around 1 hour which needs a lot of optimization, I would like to know how efficient others submissions are and number of models we might be able to fit within the 2 hour submission window.</p>",
      "rawMarkdown": "Given that this competition only allows cpu submissions which are restricted to 2 hours, its very important to build proper pipelines and made make optimize inference as much as possible, my single rough efficientnetb0 took around 1 hour which needs a lot of optimization, I would like to know how efficient others submissions are and number of models we might be able to fit within the 2 hour submission window.",
      "votes": null
    },
    {
      "id": "2744471",
      "postDate": "04/09/2024 23:39:40",
      "content": "<p>I'm using MobileNetV2 - and it's inference time is also something around 1 hour (after I got batching working correctly).</p>\n<p>I get the impression I may need to go to something a bit larger to move up the LB.</p>",
      "rawMarkdown": "I'm using MobileNetV2 - and it's inference time is also something around 1 hour (after I got batching working correctly).\n\nI get the impression I may need to go to something a bit larger to move up the LB.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2744471,
      "author_name": "richolson",
      "author_url": "",
      "post_date": "04/09/2024 23:39:40",
      "content": "<p>I'm using MobileNetV2 - and it's inference time is also something around 1 hour (after I got batching working correctly).</p>\n<p>I get the impression I may need to go to something a bit larger to move up the LB.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2742853": "Given that this competition only allows cpu submissions which are restricted to 2 hours, its very important to build proper pipelines and made make optimize inference as much as possible, my single rough efficientnetb0 took around 1 hour which needs a lot of optimization, I would like to know how efficient others submissions are and number of models we might be able to fit within the 2 hour submission window.",
    "2744471": "I'm using MobileNetV2 - and it's inference time is also something around 1 hour (after I got batching working correctly).\n\nI get the impression I may need to go to something a bit larger to move up the LB."
  },
  "source": "meta"
}