{
  "id": 491756,
  "title": "Question about runtime - Help a newcomer ",
  "url": "/competitions/birdclef-2024/discussion/491756",
  "author_name": "",
  "post_date": "2024-04-07T06:53:48.808421600Z",
  "votes": 3,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Hey everyone! It's the first competition I'm participating in, and I had a question about the runtime and CPU requirements.</p>\n<p>\"CPU Notebook &lt;= 120 minutes run-time<br>\nGPU Notebook submissions are disabled. You can technically submit but will only have 1 minute of runtime.\"</p>\n<p>I'm not sure I understand what is the runtime in question, is the model training time a part of the runtime?<br>\nIs the data processing time a part of the runtime?</p>\n<p>My second question is, am I allowed to train my model on a GPU? Can someone explain this requirement to me?</p>\n<p>Thanks for the help! </p>",
  "messages": [
    {
      "id": "2739632",
      "postDate": "04/07/2024 06:53:48",
      "content": "<p>Hey everyone! It's the first competition I'm participating in, and I had a question about the runtime and CPU requirements.</p>\n<p>\"CPU Notebook &lt;= 120 minutes run-time<br>\nGPU Notebook submissions are disabled. You can technically submit but will only have 1 minute of runtime.\"</p>\n<p>I'm not sure I understand what is the runtime in question, is the model training time a part of the runtime?<br>\nIs the data processing time a part of the runtime?</p>\n<p>My second question is, am I allowed to train my model on a GPU? Can someone explain this requirement to me?</p>\n<p>Thanks for the help! </p>",
      "rawMarkdown": "Hey everyone! It's the first competition I'm participating in, and I had a question about the runtime and CPU requirements.\n\n\"CPU Notebook <= 120 minutes run-time\nGPU Notebook submissions are disabled. You can technically submit but will only have 1 minute of runtime.\"\n\nI'm not sure I understand what is the runtime in question, is the model training time a part of the runtime?\nIs the data processing time a part of the runtime?\n\nMy second question is, am I allowed to train my model on a GPU? Can someone explain this requirement to me?\n\nThanks for the help!",
      "votes": null
    },
    {
      "id": "2739712",
      "postDate": "04/07/2024 08:05:32",
      "content": "<p>Training time is not included in runtime. \"runtime\" is the time the notebook runs when submitted. Processing on test data is also included in runtime.<br>\nAnd of course you can train your models on a GPU.</p>",
      "rawMarkdown": "Training time is not included in runtime. \"runtime\" is the time the notebook runs when submitted. Processing on test data is also included in runtime.\nAnd of course you can train your models on a GPU.",
      "votes": null
    },
    {
      "id": "2739738",
      "postDate": "04/07/2024 08:27:09",
      "content": "<p>Thanks for replying!<br>\nWill I be notified of the runtime? Or should I use python time function?</p>",
      "rawMarkdown": "Thanks for replying!\nWill I be notified of the runtime? Or should I use python time function?",
      "votes": null
    },
    {
      "id": "2739747",
      "postDate": "04/07/2024 08:33:57",
      "content": "<p>You will not be notified it. If you want to branch the process by execution time, time function is a good way to do so.</p>",
      "rawMarkdown": "You will not be notified it. If you want to branch the process by execution time, time function is a good way to do so.",
      "votes": null
    },
    {
      "id": "2739794",
      "postDate": "04/07/2024 09:44:27",
      "content": "<p>Got it, thank you and good luck!</p>",
      "rawMarkdown": "Got it, thank you and good luck!",
      "votes": null
    },
    {
      "id": "2741454",
      "postDate": "04/08/2024 11:04:55",
      "content": "<p>So far my experience is that GPU submissions will time out, but a CPU submission succeeds. I can recommend using CPU for inference, since you have 120x as much time and it will be difficult to utilize the full GPU for inference to process all 1100 samples within 1 minute.</p>",
      "rawMarkdown": "So far my experience is that GPU submissions will time out, but a CPU submission succeeds. I can recommend using CPU for inference, since you have 120x as much time and it will be difficult to utilize the full GPU for inference to process all 1100 samples within 1 minute.",
      "votes": null
    },
    {
      "id": "2741790",
      "postDate": "04/08/2024 15:44:39",
      "content": "<p>Thanks for the clarification! appreciate it</p>",
      "rawMarkdown": "Thanks for the clarification! appreciate it",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2739712,
      "author_name": "ryotayoshinobu",
      "author_url": "",
      "post_date": "04/07/2024 08:05:32",
      "content": "<p>Training time is not included in runtime. \"runtime\" is the time the notebook runs when submitted. Processing on test data is also included in runtime.<br>\nAnd of course you can train your models on a GPU.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2739738,
          "author_name": "guyinbar33",
          "author_url": "",
          "post_date": "04/07/2024 08:27:09",
          "content": "<p>Thanks for replying!<br>\nWill I be notified of the runtime? Or should I use python time function?</p>",
          "votes": null,
          "replies": [
            {
              "id": 2739747,
              "author_name": "ryotayoshinobu",
              "author_url": "",
              "post_date": "04/07/2024 08:33:57",
              "content": "<p>You will not be notified it. If you want to branch the process by execution time, time function is a good way to do so.</p>",
              "votes": null,
              "replies": [
                {
                  "id": 2739794,
                  "author_name": "guyinbar33",
                  "author_url": "",
                  "post_date": "04/07/2024 09:44:27",
                  "content": "<p>Got it, thank you and good luck!</p>",
                  "votes": null,
                  "replies": []
                }
              ]
            }
          ]
        }
      ]
    },
    {
      "id": 2741454,
      "author_name": "markwijkhuizen",
      "author_url": "",
      "post_date": "04/08/2024 11:04:55",
      "content": "<p>So far my experience is that GPU submissions will time out, but a CPU submission succeeds. I can recommend using CPU for inference, since you have 120x as much time and it will be difficult to utilize the full GPU for inference to process all 1100 samples within 1 minute.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2741790,
          "author_name": "guyinbar33",
          "author_url": "",
          "post_date": "04/08/2024 15:44:39",
          "content": "<p>Thanks for the clarification! appreciate it</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2739632": "Hey everyone! It's the first competition I'm participating in, and I had a question about the runtime and CPU requirements.\n\n\"CPU Notebook <= 120 minutes run-time\nGPU Notebook submissions are disabled. You can technically submit but will only have 1 minute of runtime.\"\n\nI'm not sure I understand what is the runtime in question, is the model training time a part of the runtime?\nIs the data processing time a part of the runtime?\n\nMy second question is, am I allowed to train my model on a GPU? Can someone explain this requirement to me?\n\nThanks for the help!",
    "2739712": "Training time is not included in runtime. \"runtime\" is the time the notebook runs when submitted. Processing on test data is also included in runtime.\nAnd of course you can train your models on a GPU.",
    "2739738": "Thanks for replying!\nWill I be notified of the runtime? Or should I use python time function?",
    "2739747": "You will not be notified it. If you want to branch the process by execution time, time function is a good way to do so.",
    "2739794": "Got it, thank you and good luck!",
    "2741454": "So far my experience is that GPU submissions will time out, but a CPU submission succeeds. I can recommend using CPU for inference, since you have 120x as much time and it will be difficult to utilize the full GPU for inference to process all 1100 samples within 1 minute.",
    "2741790": "Thanks for the clarification! appreciate it"
  },
  "source": "meta"
}