{
  "id": 572617,
  "title": "Submission Takes Long Time",
  "url": "/competitions/byu-locating-bacterial-flagellar-motors-2025/discussion/572617",
  "author_name": "",
  "post_date": "2025-04-10T13:08:36.230351200Z",
  "votes": 2,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Why submitting to this competition takes a long time?</p>",
  "messages": [
    {
      "id": "3175665",
      "postDate": "04/10/2025 13:08:36",
      "content": "<p>Why submitting to this competition takes a long time?</p>",
      "rawMarkdown": "Why submitting to this competition takes a long time?",
      "votes": null
    },
    {
      "id": "3175699",
      "postDate": "04/10/2025 13:45:05",
      "content": "<p>Yes of course, this is influenced by your code, if you train in the submission kernel and the choice of your model for inferencing <a href=\"https://www.kaggle.com/meshkatahammed\" target=\"_blank\">@meshkatahammed</a> <br>\nPlease be mindful of the submission time in case you wish to submit a long code. You could also choose to train and infer separately </p>",
      "rawMarkdown": "Yes of course, this is influenced by your code, if you train in the submission kernel and the choice of your model for inferencing @meshkatahammed \nPlease be mindful of the submission time in case you wish to submit a long code. You could also choose to train and infer separately",
      "votes": null
    },
    {
      "id": "3176106",
      "postDate": "04/11/2025 01:40:38",
      "content": "<p>I am training and inferencing separately. Model weight is only being used in the submission notebook. Using GPU P100. Still it takes more than 9 hours to get submitted. Please suggest any hack. Thanks!</p>",
      "rawMarkdown": "I am training and inferencing separately. Model weight is only being used in the submission notebook. Using GPU P100. Still it takes more than 9 hours to get submitted. Please suggest any hack. Thanks!",
      "votes": null
    },
    {
      "id": "3176433",
      "postDate": "04/11/2025 10:45:38",
      "content": "<p>The test set is large and it takes time to read the slices. That said 9h is pretty long. You can try to run your submission notebook on the train set to see where you take to much time. There are some ways to speed up computation like TensorRT and using T4 x2 GPUs.</p>",
      "rawMarkdown": "The test set is large and it takes time to read the slices. That said 9h is pretty long. You can try to run your submission notebook on the train set to see where you take to much time. There are some ways to speed up computation like TensorRT and using T4 x2 GPUs.",
      "votes": null
    },
    {
      "id": "3176857",
      "postDate": "04/11/2025 19:58:12",
      "content": "<p>You can check my optimized submission notebook, and try to adopt it for your model/pipeline <a href=\"https://www.kaggle.com/code/fautei/byu-yolo-optimized-submission-notebook\" target=\"_blank\">https://www.kaggle.com/code/fautei/byu-yolo-optimized-submission-notebook</a></p>",
      "rawMarkdown": "You can check my optimized submission notebook, and try to adopt it for your model/pipeline https://www.kaggle.com/code/fautei/byu-yolo-optimized-submission-notebook",
      "votes": null
    },
    {
      "id": "3176952",
      "postDate": "04/12/2025 02:14:01",
      "content": "<p>When I use T4 x 2 GPUs it actually utilizes only one GPU, the other one stays idle. That's why I have decided to use P100. </p>",
      "rawMarkdown": "When I use T4 x 2 GPUs it actually utilizes only one GPU, the other one stays idle. That's why I have decided to use P100.",
      "votes": null
    },
    {
      "id": "3178193",
      "postDate": "04/13/2025 20:26:36",
      "content": "<p>Yes, it takes a lot of time — almost as much as the execution time of your code. That's because Kaggle runs and stores the code on its own servers. The first time, it runs your code, and then it executes the code on the competition data for evaluation.</p>",
      "rawMarkdown": "Yes, it takes a lot of time — almost as much as the execution time of your code. That's because Kaggle runs and stores the code on its own servers. The first time, it runs your code, and then it executes the code on the competition data for evaluation.",
      "votes": null
    },
    {
      "id": "3178285",
      "postDate": "04/14/2025 01:56:06",
      "content": "<p>But still my last submission has even timed out. I wonder if the submissions wait in the queue and get executed one by one. Is there any hack that I can apply to accelerate submission?</p>",
      "rawMarkdown": "But still my last submission has even timed out. I wonder if the submissions wait in the queue and get executed one by one. Is there any hack that I can apply to accelerate submission?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3175699,
      "author_name": "ravi20076",
      "author_url": "",
      "post_date": "04/10/2025 13:45:05",
      "content": "<p>Yes of course, this is influenced by your code, if you train in the submission kernel and the choice of your model for inferencing <a href=\"https://www.kaggle.com/meshkatahammed\" target=\"_blank\">@meshkatahammed</a> <br>\nPlease be mindful of the submission time in case you wish to submit a long code. You could also choose to train and infer separately </p>",
      "votes": null,
      "replies": [
        {
          "id": 3176106,
          "author_name": "meshkatahammed",
          "author_url": "",
          "post_date": "04/11/2025 01:40:38",
          "content": "<p>I am training and inferencing separately. Model weight is only being used in the submission notebook. Using GPU P100. Still it takes more than 9 hours to get submitted. Please suggest any hack. Thanks!</p>",
          "votes": null,
          "replies": [
            {
              "id": 3176857,
              "author_name": "fautei",
              "author_url": "",
              "post_date": "04/11/2025 19:58:12",
              "content": "<p>You can check my optimized submission notebook, and try to adopt it for your model/pipeline <a href=\"https://www.kaggle.com/code/fautei/byu-yolo-optimized-submission-notebook\" target=\"_blank\">https://www.kaggle.com/code/fautei/byu-yolo-optimized-submission-notebook</a></p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 3176433,
      "author_name": "tennogh",
      "author_url": "",
      "post_date": "04/11/2025 10:45:38",
      "content": "<p>The test set is large and it takes time to read the slices. That said 9h is pretty long. You can try to run your submission notebook on the train set to see where you take to much time. There are some ways to speed up computation like TensorRT and using T4 x2 GPUs.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3176952,
          "author_name": "meshkatahammed",
          "author_url": "",
          "post_date": "04/12/2025 02:14:01",
          "content": "<p>When I use T4 x 2 GPUs it actually utilizes only one GPU, the other one stays idle. That's why I have decided to use P100. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 3178193,
      "author_name": "ali78kabirzadeh",
      "author_url": "",
      "post_date": "04/13/2025 20:26:36",
      "content": "<p>Yes, it takes a lot of time — almost as much as the execution time of your code. That's because Kaggle runs and stores the code on its own servers. The first time, it runs your code, and then it executes the code on the competition data for evaluation.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3178285,
          "author_name": "meshkatahammed",
          "author_url": "",
          "post_date": "04/14/2025 01:56:06",
          "content": "<p>But still my last submission has even timed out. I wonder if the submissions wait in the queue and get executed one by one. Is there any hack that I can apply to accelerate submission?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3175665": "Why submitting to this competition takes a long time?",
    "3175699": "Yes of course, this is influenced by your code, if you train in the submission kernel and the choice of your model for inferencing @meshkatahammed \nPlease be mindful of the submission time in case you wish to submit a long code. You could also choose to train and infer separately",
    "3176106": "I am training and inferencing separately. Model weight is only being used in the submission notebook. Using GPU P100. Still it takes more than 9 hours to get submitted. Please suggest any hack. Thanks!",
    "3176433": "The test set is large and it takes time to read the slices. That said 9h is pretty long. You can try to run your submission notebook on the train set to see where you take to much time. There are some ways to speed up computation like TensorRT and using T4 x2 GPUs.",
    "3176857": "You can check my optimized submission notebook, and try to adopt it for your model/pipeline https://www.kaggle.com/code/fautei/byu-yolo-optimized-submission-notebook",
    "3176952": "When I use T4 x 2 GPUs it actually utilizes only one GPU, the other one stays idle. That's why I have decided to use P100.",
    "3178193": "Yes, it takes a lot of time — almost as much as the execution time of your code. That's because Kaggle runs and stores the code on its own servers. The first time, it runs your code, and then it executes the code on the competition data for evaluation.",
    "3178285": "But still my last submission has even timed out. I wonder if the submissions wait in the queue and get executed one by one. Is there any hack that I can apply to accelerate submission?"
  },
  "source": "meta"
}