{
  "id": 412861,
  "title": "New to 'Code Competition', Can anyone explain some terms?",
  "url": "/competitions/asl-fingerspelling/discussion/412861",
  "author_name": "Andrew",
  "post_date": "2023-05-25T14:24:10.625000",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello, I have some questions,</p>\n<ol>\n<li>Does 'Code Competiton' means that all model must be trained in the Notebook? </li>\n<li>Does the Notebook Runtime is total amount of training+testing or just in training phase?<br>\nCPU Notebook &lt;= 9 hours run-time<br>\nGPU Notebook &lt;= 9 hours run-time</li>\n<li>Does 'Internet access disabled' mean we can't use external models? like LSTM, RNN, CNN? Or I can still import RNN? What allowed and what not? But they said that freely available data is allowed? <br>\nThanks!</li>\n</ol>",
  "messages": [
    {
      "id": 2274047,
      "postDate": "2023-05-25T15:24:59.050Z",
      "content": "<p>Hi, <a href=\"https://www.kaggle.com/andrewfirman\" target=\"_blank\">@andrewfirman</a>. </p>\n<ol>\n<li><p>You could train the model anywhere you want - locally, in the cloud, in google colab, in kaggle, you named it. You should submit a notebook that produces a model. You can train it from scratch, but you also can also upload it to the kaggle datasets and attach it to the notebook. My notebook contains only a <code>!cp model ./dst</code>-like commands.</p></li>\n<li><p>It is a general requirement. In this particular competition, we faced more strict requirements: Your model must also perform inference in less than 5 hours and use less than 40 MB of storage space. Expect to see approximately 35 hours of video in the test set.</p></li>\n<li><p>You could bypass anything you want via kaggle datasets. Upload the equired files to the kaggle datasets and attach this dataset to your notebook. This data will be available during submission. It may also include 3rd party libraries to install (like pip wheels, etc).</p></li>\n</ol>",
      "rawMarkdown": "Hi, @andrewfirman. \n\n1. You could train the model anywhere you want - locally, in the cloud, in google colab, in kaggle, you named it. You should submit a notebook that produces a model. You can train it from scratch, but you also can also upload it to the kaggle datasets and attach it to the notebook. My notebook contains only a `!cp model ./dst`-like commands.\n\n2. It is a general requirement. In this particular competition, we faced more strict requirements: Your model must also perform inference in less than 5 hours and use less than 40 MB of storage space. Expect to see approximately 35 hours of video in the test set.\n\n3. You could bypass anything you want via kaggle datasets. Upload the equired files to the kaggle datasets and attach this dataset to your notebook. This data will be available during submission. It may also include 3rd party libraries to install (like pip wheels, etc).",
      "votes": 2,
      "replies": [
        {
          "id": 2274463,
          "postDate": "2023-05-26T03:09:08.050Z",
          "content": "<p>Thank you!</p>",
          "rawMarkdown": "Thank you!"
        }
      ]
    },
    {
      "id": 2273973,
      "postDate": "2023-05-25T14:24:10.627Z",
      "content": "<p>Hello, I have some questions,</p>\n<ol>\n<li>Does 'Code Competiton' means that all model must be trained in the Notebook? </li>\n<li>Does the Notebook Runtime is total amount of training+testing or just in training phase?<br>\nCPU Notebook &lt;= 9 hours run-time<br>\nGPU Notebook &lt;= 9 hours run-time</li>\n<li>Does 'Internet access disabled' mean we can't use external models? like LSTM, RNN, CNN? Or I can still import RNN? What allowed and what not? But they said that freely available data is allowed? <br>\nThanks!</li>\n</ol>",
      "rawMarkdown": "Hello, I have some questions,\n1. Does 'Code Competiton' means that all model must be trained in the Notebook? \n2. Does the Notebook Runtime is total amount of training+testing or just in training phase?\n    CPU Notebook <= 9 hours run-time\n    GPU Notebook <= 9 hours run-time\n3. Does 'Internet access disabled' mean we can't use external models? like LSTM, RNN, CNN? Or I can still import RNN? What allowed and what not? But they said that freely available data is allowed? \n    Thanks!\n",
      "votes": 2
    }
  ],
  "comments": [
    {
      "id": 2274047,
      "author_name": "Mykola",
      "author_url": "",
      "post_date": "2023-05-25T15:24:59.050000",
      "content": "<p>Hi, <a href=\"https://www.kaggle.com/andrewfirman\" target=\"_blank\">@andrewfirman</a>. </p>\n<ol>\n<li><p>You could train the model anywhere you want - locally, in the cloud, in google colab, in kaggle, you named it. You should submit a notebook that produces a model. You can train it from scratch, but you also can also upload it to the kaggle datasets and attach it to the notebook. My notebook contains only a <code>!cp model ./dst</code>-like commands.</p></li>\n<li><p>It is a general requirement. In this particular competition, we faced more strict requirements: Your model must also perform inference in less than 5 hours and use less than 40 MB of storage space. Expect to see approximately 35 hours of video in the test set.</p></li>\n<li><p>You could bypass anything you want via kaggle datasets. Upload the equired files to the kaggle datasets and attach this dataset to your notebook. This data will be available during submission. It may also include 3rd party libraries to install (like pip wheels, etc).</p></li>\n</ol>",
      "votes": 2,
      "replies": [
        {
          "id": 2274463,
          "author_name": "Andrew",
          "author_url": "",
          "post_date": "2023-05-26T03:09:08.050000",
          "content": "<p>Thank you!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2274047": "Hi, @andrewfirman. \n\n1. You could train the model anywhere you want - locally, in the cloud, in google colab, in kaggle, you named it. You should submit a notebook that produces a model. You can train it from scratch, but you also can also upload it to the kaggle datasets and attach it to the notebook. My notebook contains only a `!cp model ./dst`-like commands.\n\n2. It is a general requirement. In this particular competition, we faced more strict requirements: Your model must also perform inference in less than 5 hours and use less than 40 MB of storage space. Expect to see approximately 35 hours of video in the test set.\n\n3. You could bypass anything you want via kaggle datasets. Upload the equired files to the kaggle datasets and attach this dataset to your notebook. This data will be available during submission. It may also include 3rd party libraries to install (like pip wheels, etc).",
    "2273973": "Hello, I have some questions,\n1. Does 'Code Competiton' means that all model must be trained in the Notebook? \n2. Does the Notebook Runtime is total amount of training+testing or just in training phase?\n    CPU Notebook <= 9 hours run-time\n    GPU Notebook <= 9 hours run-time\n3. Does 'Internet access disabled' mean we can't use external models? like LSTM, RNN, CNN? Or I can still import RNN? What allowed and what not? But they said that freely available data is allowed? \n    Thanks!\n"
  }
}