{
  "id": 350207,
  "title": "showing submission running time.",
  "url": "/competitions/rsna-2022-cervical-spine-fracture-detection/discussion/350207",
  "author_name": "",
  "post_date": "2022-09-04T17:05:10.863094900Z",
  "votes": 4,
  "comment_count": 5,
  "views": 0,
  "content": "<p>instead of only the score, putting the inference time aside is helpful.</p>",
  "messages": [
    {
      "id": "1926254",
      "postDate": "09/04/2022 17:05:10",
      "content": "<p>instead of only the score, putting the inference time aside is helpful.</p>",
      "rawMarkdown": "instead of only the score, putting the inference time aside is helpful.",
      "votes": null
    },
    {
      "id": "1926495",
      "postDate": "09/04/2022 22:14:36",
      "content": "<p>Nice tip!, Updated!</p>\n<p>5fold resnext50_32x4d - 512x512 LB: 0.28, [POST] [RUNTIME: 7-8 HOURS]<br>\n5fold resnext50_32x4d - 512x512 LB: 0.31, [FULL DATA] [RUNTIME: 2.5 HOURS]<br>\n5fold resnext50_32x4d - 512x512 LB: 0.37 [PART DATA] [RUNTIME: 2.5 HOURS]</p>",
      "rawMarkdown": "Nice tip!, Updated!\n\n5fold resnext50_32x4d - 512x512 LB: 0.28, [POST] [RUNTIME: 7-8 HOURS]\n5fold resnext50_32x4d - 512x512 LB: 0.31, [FULL DATA] [RUNTIME: 2.5 HOURS]\n5fold resnext50_32x4d - 512x512 LB: 0.37 [PART DATA] [RUNTIME: 2.5 HOURS]",
      "votes": null
    },
    {
      "id": "1928258",
      "postDate": "09/06/2022 11:18:18",
      "content": "<p>runtime: about 3h  img_size=224*224</p>",
      "rawMarkdown": "runtime: about 3h  img_size=224*224",
      "votes": null
    },
    {
      "id": "1942373",
      "postDate": "09/16/2022 15:54:59",
      "content": "<p>I just realize that this is very good runtime. Our 5fold resnext50_32x4d takes 6 hours to run using all slices. Did you find a good way to optimize the network inference or are you treating the data differently? </p>",
      "rawMarkdown": "I just realize that this is very good runtime. Our 5fold resnext50_32x4d takes 6 hours to run using all slices. Did you find a good way to optimize the network inference or are you treating the data differently?",
      "votes": null
    },
    {
      "id": "1942432",
      "postDate": "09/16/2022 16:35:35",
      "content": "<p>Nope, I am running it nowhere near all slices, my version you could say is very low optimized (because I don't have the need currently, will surely need it later, and is already planned how to achieve), it is about 3x7 = 21 images per patient</p>",
      "rawMarkdown": "Nope, I am running it nowhere near all slices, my version you could say is very low optimized (because I don't have the need currently, will surely need it later, and is already planned how to achieve), it is about 3x7 = 21 images per patient",
      "votes": null
    },
    {
      "id": "1942613",
      "postDate": "09/16/2022 19:19:04",
      "content": "<p>Thanks for sharing it. I didn't find a good way to aggregate slices to use fewer images yet but I'll certainly try that. </p>",
      "rawMarkdown": "Thanks for sharing it. I didn't find a good way to aggregate slices to use fewer images yet but I'll certainly try that.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1926495,
      "author_name": "harshitsheoran",
      "author_url": "",
      "post_date": "09/04/2022 22:14:36",
      "content": "<p>Nice tip!, Updated!</p>\n<p>5fold resnext50_32x4d - 512x512 LB: 0.28, [POST] [RUNTIME: 7-8 HOURS]<br>\n5fold resnext50_32x4d - 512x512 LB: 0.31, [FULL DATA] [RUNTIME: 2.5 HOURS]<br>\n5fold resnext50_32x4d - 512x512 LB: 0.37 [PART DATA] [RUNTIME: 2.5 HOURS]</p>",
      "votes": null,
      "replies": [
        {
          "id": 1942373,
          "author_name": "igormunizims",
          "author_url": "",
          "post_date": "09/16/2022 15:54:59",
          "content": "<p>I just realize that this is very good runtime. Our 5fold resnext50_32x4d takes 6 hours to run using all slices. Did you find a good way to optimize the network inference or are you treating the data differently? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1942432,
          "author_name": "harshitsheoran",
          "author_url": "",
          "post_date": "09/16/2022 16:35:35",
          "content": "<p>Nope, I am running it nowhere near all slices, my version you could say is very low optimized (because I don't have the need currently, will surely need it later, and is already planned how to achieve), it is about 3x7 = 21 images per patient</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1942613,
          "author_name": "igormunizims",
          "author_url": "",
          "post_date": "09/16/2022 19:19:04",
          "content": "<p>Thanks for sharing it. I didn't find a good way to aggregate slices to use fewer images yet but I'll certainly try that. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1928258,
      "author_name": "abebe9849",
      "author_url": "",
      "post_date": "09/06/2022 11:18:18",
      "content": "<p>runtime: about 3h  img_size=224*224</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1926254": "instead of only the score, putting the inference time aside is helpful.",
    "1926495": "Nice tip!, Updated!\n\n5fold resnext50_32x4d - 512x512 LB: 0.28, [POST] [RUNTIME: 7-8 HOURS]\n5fold resnext50_32x4d - 512x512 LB: 0.31, [FULL DATA] [RUNTIME: 2.5 HOURS]\n5fold resnext50_32x4d - 512x512 LB: 0.37 [PART DATA] [RUNTIME: 2.5 HOURS]",
    "1928258": "runtime: about 3h  img_size=224*224",
    "1942373": "I just realize that this is very good runtime. Our 5fold resnext50_32x4d takes 6 hours to run using all slices. Did you find a good way to optimize the network inference or are you treating the data differently?",
    "1942432": "Nope, I am running it nowhere near all slices, my version you could say is very low optimized (because I don't have the need currently, will surely need it later, and is already planned how to achieve), it is about 3x7 = 21 images per patient",
    "1942613": "Thanks for sharing it. I didn't find a good way to aggregate slices to use fewer images yet but I'll certainly try that."
  },
  "source": "meta"
}