{
  "id": 173264,
  "title": "Code requirements - Question",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/173264",
  "author_name": "",
  "post_date": "2020-08-08T14:30:14.147359800Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>I am new on Kaggle and I am a bit confused by below requirements.</p>\n<p>*&gt; Submissions to this competition must be made through Notebooks. In order for the \"Submit to Competition\" button to be active after a commit, the following conditions must be met:</p>\n<p>CPU Notebook &lt;= 9 hours run-time<br>\nGPU Notebook &lt;= 4 hours run-time<br>\nTPUs will not be available for making submissions to this competition. You are still welcome to use them for training models.<br>\nNo internet access enabled<br>\nExternal data, freely &amp; publicly available, is allowed. This includes pre-trained models.<br>\nSubmission file must be named submission.csv*</p>\n<p>What does it mean that submissions must be made using Notebooks only? <br>\nCan I not use my personal GPU to train models??<br>\nCan we train model on our own setup and then use the weights to do test predictions and then do the submissions through notebook??<br>\nWhat does CPU and GPU run time mean?? Is it the runtime for processing and prediction??<br>\nOr is it total run time needed to train a model and do the predictions?</p>\n<p>Please clarify this.</p>",
  "messages": [
    {
      "id": "962904",
      "postDate": "08/08/2020 14:30:14",
      "content": "<p>Hi,</p>\n<p>I am new on Kaggle and I am a bit confused by below requirements.</p>\n<p>*&gt; Submissions to this competition must be made through Notebooks. In order for the \"Submit to Competition\" button to be active after a commit, the following conditions must be met:</p>\n<p>CPU Notebook &lt;= 9 hours run-time<br>\nGPU Notebook &lt;= 4 hours run-time<br>\nTPUs will not be available for making submissions to this competition. You are still welcome to use them for training models.<br>\nNo internet access enabled<br>\nExternal data, freely &amp; publicly available, is allowed. This includes pre-trained models.<br>\nSubmission file must be named submission.csv*</p>\n<p>What does it mean that submissions must be made using Notebooks only? <br>\nCan I not use my personal GPU to train models??<br>\nCan we train model on our own setup and then use the weights to do test predictions and then do the submissions through notebook??<br>\nWhat does CPU and GPU run time mean?? Is it the runtime for processing and prediction??<br>\nOr is it total run time needed to train a model and do the predictions?</p>\n<p>Please clarify this.</p>",
      "rawMarkdown": "Hi,\n\nI am new on Kaggle and I am a bit confused by below requirements.\n\n*&gt; Submissions to this competition must be made through Notebooks. In order for the \"Submit to Competition\" button to be active after a commit, the following conditions must be met:\n\nCPU Notebook &lt;= 9 hours run-time\nGPU Notebook &lt;= 4 hours run-time\nTPUs will not be available for making submissions to this competition. You are still welcome to use them for training models.\nNo internet access enabled\nExternal data, freely &amp; publicly available, is allowed. This includes pre-trained models.\nSubmission file must be named submission.csv*\n\nWhat does it mean that submissions must be made using Notebooks only? \nCan I not use my personal GPU to train models??\nCan we train model on our own setup and then use the weights to do test predictions and then do the submissions through notebook??\nWhat does CPU and GPU run time mean?? Is it the runtime for processing and prediction??\nOr is it total run time needed to train a model and do the predictions?\n\nPlease clarify this.",
      "votes": null
    },
    {
      "id": "962986",
      "postDate": "08/08/2020 15:28:03",
      "content": "<p>The limitations are on the final submission. You can train locally. You can preprocess the test data locally for local training. You can upload your model.</p>\n\n<p>Then you must submit a notebook that will in batch mode read the real test.csv and test dicom data and produce a submission.csv file. This will include prediction on the test data.</p>\n\n<p>So, if you preprocess and produce a model and upload it, the CPU/GPU/Time limits are only on prediction. If you do everything from scratch in your notebook, the time limits would include preprocessing, training and prediction.</p>\n\n<p>You cannot submit a submission.csv file directly (and you cannot because you don't have access to the real test data). You must submit a notebook that creates a submission.csv file on the fly.</p>",
      "rawMarkdown": "The limitations are on the final submission. You can train locally. You can preprocess the test data locally for local training. You can upload your model.\n\nThen you must submit a notebook that will in batch mode read the real test.csv and test dicom data and produce a submission.csv file. This will include prediction on the test data.\n\nSo, if you preprocess and produce a model and upload it, the CPU/GPU/Time limits are only on prediction. If you do everything from scratch in your notebook, the time limits would include preprocessing, training and prediction.\n\nYou cannot submit a submission.csv file directly (and you cannot because you don't have access to the real test data). You must submit a notebook that creates a submission.csv file on the fly.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 962986,
      "author_name": "richardepstein",
      "author_url": "",
      "post_date": "08/08/2020 15:28:03",
      "content": "<p>The limitations are on the final submission. You can train locally. You can preprocess the test data locally for local training. You can upload your model.</p>\n\n<p>Then you must submit a notebook that will in batch mode read the real test.csv and test dicom data and produce a submission.csv file. This will include prediction on the test data.</p>\n\n<p>So, if you preprocess and produce a model and upload it, the CPU/GPU/Time limits are only on prediction. If you do everything from scratch in your notebook, the time limits would include preprocessing, training and prediction.</p>\n\n<p>You cannot submit a submission.csv file directly (and you cannot because you don't have access to the real test data). You must submit a notebook that creates a submission.csv file on the fly.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "962904": "Hi,\n\nI am new on Kaggle and I am a bit confused by below requirements.\n\n*&gt; Submissions to this competition must be made through Notebooks. In order for the \"Submit to Competition\" button to be active after a commit, the following conditions must be met:\n\nCPU Notebook &lt;= 9 hours run-time\nGPU Notebook &lt;= 4 hours run-time\nTPUs will not be available for making submissions to this competition. You are still welcome to use them for training models.\nNo internet access enabled\nExternal data, freely &amp; publicly available, is allowed. This includes pre-trained models.\nSubmission file must be named submission.csv*\n\nWhat does it mean that submissions must be made using Notebooks only? \nCan I not use my personal GPU to train models??\nCan we train model on our own setup and then use the weights to do test predictions and then do the submissions through notebook??\nWhat does CPU and GPU run time mean?? Is it the runtime for processing and prediction??\nOr is it total run time needed to train a model and do the predictions?\n\nPlease clarify this.",
    "962986": "The limitations are on the final submission. You can train locally. You can preprocess the test data locally for local training. You can upload your model.\n\nThen you must submit a notebook that will in batch mode read the real test.csv and test dicom data and produce a submission.csv file. This will include prediction on the test data.\n\nSo, if you preprocess and produce a model and upload it, the CPU/GPU/Time limits are only on prediction. If you do everything from scratch in your notebook, the time limits would include preprocessing, training and prediction.\n\nYou cannot submit a submission.csv file directly (and you cannot because you don't have access to the real test data). You must submit a notebook that creates a submission.csv file on the fly."
  },
  "source": "meta"
}