{
  "id": 401248,
  "title": "Code Competition Run-time Explanation",
  "url": "/competitions/benetech-making-graphs-accessible/discussion/401248",
  "author_name": "",
  "post_date": "2023-04-12T13:23:27.104639500Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Hi everyone, apologies for the basic questions. I am new to the code competition format and I have a question about the rules here.</p>\n<p>The competition's rules stated these two things:</p>\n<blockquote>\n  <p>CPU Notebook &lt;= 9 hours run-time<br>\n  GPU Notebook &lt;= 9 hours run-time</p>\n</blockquote>\n<p>Does the 9 hours time limit include the training part or just the inference on the test set?</p>\n<p>Thank you.</p>",
  "messages": [
    {
      "id": "2219296",
      "postDate": "04/12/2023 13:23:27",
      "content": "<p>Hi everyone, apologies for the basic questions. I am new to the code competition format and I have a question about the rules here.</p>\n<p>The competition's rules stated these two things:</p>\n<blockquote>\n  <p>CPU Notebook &lt;= 9 hours run-time<br>\n  GPU Notebook &lt;= 9 hours run-time</p>\n</blockquote>\n<p>Does the 9 hours time limit include the training part or just the inference on the test set?</p>\n<p>Thank you.</p>",
      "rawMarkdown": "Hi everyone, apologies for the basic questions. I am new to the code competition format and I have a question about the rules here.\n\nThe competition's rules stated these two things:\n\n>CPU Notebook <= 9 hours run-time\nGPU Notebook <= 9 hours run-time\n\nDoes the 9 hours time limit include the training part or just the inference on the test set?\n\nThank you.",
      "votes": null
    },
    {
      "id": "2219647",
      "postDate": "04/12/2023 18:28:47",
      "content": "<p>It is hard to do both under 9 hours. <br>\nMost of people do training and inference in separate notebooks. <br>\nAs long as the inference time is below 9 hours, you will be fine. </p>",
      "rawMarkdown": "It is hard to do both under 9 hours. \nMost of people do training and inference in separate notebooks. \nAs long as the inference time is below 9 hours, you will be fine.",
      "votes": null
    },
    {
      "id": "2219767",
      "postDate": "04/12/2023 21:10:42",
      "content": "<p>Thank you for your reply. Does that mean people usually train on one notebook -&gt; save the trained model -&gt; load the trained model on the new notebook to perform inference?</p>",
      "rawMarkdown": "Thank you for your reply. Does that mean people usually train on one notebook -> save the trained model -> load the trained model on the new notebook to perform inference?",
      "votes": null
    },
    {
      "id": "2219810",
      "postDate": "04/12/2023 22:30:45",
      "content": "<p>Yes <br>\nFor example , I usually train using colab, then load the trained models here and perform inference. </p>",
      "rawMarkdown": "Yes \nFor example , I usually train using colab, then load the trained models here and perform inference.",
      "votes": null
    },
    {
      "id": "2220922",
      "postDate": "04/13/2023 19:22:36",
      "content": "<p>Could we use our own GPU offline to train the model and then upload the result to perform inference on Kaggle? Thanks in advance.</p>",
      "rawMarkdown": "Could we use our own GPU offline to train the model and then upload the result to perform inference on Kaggle? Thanks in advance.",
      "votes": null
    },
    {
      "id": "2220960",
      "postDate": "04/13/2023 20:15:45",
      "content": "<p>You sure can do that!</p>",
      "rawMarkdown": "You sure can do that!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2219647,
      "author_name": "zhugds",
      "author_url": "",
      "post_date": "04/12/2023 18:28:47",
      "content": "<p>It is hard to do both under 9 hours. <br>\nMost of people do training and inference in separate notebooks. <br>\nAs long as the inference time is below 9 hours, you will be fine. </p>",
      "votes": null,
      "replies": [
        {
          "id": 2219767,
          "author_name": "mufathurrohman",
          "author_url": "",
          "post_date": "04/12/2023 21:10:42",
          "content": "<p>Thank you for your reply. Does that mean people usually train on one notebook -&gt; save the trained model -&gt; load the trained model on the new notebook to perform inference?</p>",
          "votes": null,
          "replies": [
            {
              "id": 2219810,
              "author_name": "asalhi",
              "author_url": "",
              "post_date": "04/12/2023 22:30:45",
              "content": "<p>Yes <br>\nFor example , I usually train using colab, then load the trained models here and perform inference. </p>",
              "votes": null,
              "replies": [
                {
                  "id": 2220922,
                  "author_name": "zhangzmkevin",
                  "author_url": "",
                  "post_date": "04/13/2023 19:22:36",
                  "content": "<p>Could we use our own GPU offline to train the model and then upload the result to perform inference on Kaggle? Thanks in advance.</p>",
                  "votes": null,
                  "replies": [
                    {
                      "id": 2220960,
                      "author_name": "zhugds",
                      "author_url": "",
                      "post_date": "04/13/2023 20:15:45",
                      "content": "<p>You sure can do that!</p>",
                      "votes": null,
                      "replies": []
                    }
                  ]
                }
              ]
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2219296": "Hi everyone, apologies for the basic questions. I am new to the code competition format and I have a question about the rules here.\n\nThe competition's rules stated these two things:\n\n>CPU Notebook <= 9 hours run-time\nGPU Notebook <= 9 hours run-time\n\nDoes the 9 hours time limit include the training part or just the inference on the test set?\n\nThank you.",
    "2219647": "It is hard to do both under 9 hours. \nMost of people do training and inference in separate notebooks. \nAs long as the inference time is below 9 hours, you will be fine.",
    "2219767": "Thank you for your reply. Does that mean people usually train on one notebook -> save the trained model -> load the trained model on the new notebook to perform inference?",
    "2219810": "Yes \nFor example , I usually train using colab, then load the trained models here and perform inference.",
    "2220922": "Could we use our own GPU offline to train the model and then upload the result to perform inference on Kaggle? Thanks in advance.",
    "2220960": "You sure can do that!"
  },
  "source": "meta"
}