{
  "id": 42981,
  "title": "Is there a required format of model for submission?",
  "url": "/competitions/passenger-screening-algorithm-challenge/discussion/42981",
  "author_name": "",
  "post_date": "2017-11-08T05:40:38.412245Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi,\nSorry, this is a newbie question, but\nWe have to submit our model code at the very latest on December 10th. </p>\n\n<p>1) Is there any requirement in terms of how the code should be structured? </p>\n\n<p>I am asking because my understanding that kaggle will verify the validity of our stage 2 label submission by running our code themselves against stage 2 test data. Say my code is written in a way that leverages third party cloud environment. I suppose I need to provide an instruction on how to deploy the code and test data to run on that environment?</p>\n\n<p>2) we need to submit Pre-trained models by December 4th. What exactly are \"pre-trained models\"?  Does it simply mean December 4th is the last day we can submit stage 1 test data result to check our model performance?</p>\n\n<p>thanks!</p>",
  "messages": [
    {
      "id": "241112",
      "postDate": "11/08/2017 05:40:38",
      "content": "<p>Hi,\nSorry, this is a newbie question, but\nWe have to submit our model code at the very latest on December 10th. </p>\n\n<p>1) Is there any requirement in terms of how the code should be structured? </p>\n\n<p>I am asking because my understanding that kaggle will verify the validity of our stage 2 label submission by running our code themselves against stage 2 test data. Say my code is written in a way that leverages third party cloud environment. I suppose I need to provide an instruction on how to deploy the code and test data to run on that environment?</p>\n\n<p>2) we need to submit Pre-trained models by December 4th. What exactly are \"pre-trained models\"?  Does it simply mean December 4th is the last day we can submit stage 1 test data result to check our model performance?</p>\n\n<p>thanks!</p>",
      "rawMarkdown": "Hi,\nSorry, this is a newbie question, but\nWe have to submit our model code at the very latest on December 10th. \n\n1) Is there any requirement in terms of how the code should be structured? \n\nI am asking because my understanding that kaggle will verify the validity of our stage 2 label submission by running our code themselves against stage 2 test data. Say my code is written in a way that leverages third party cloud environment. I suppose I need to provide an instruction on how to deploy the code and test data to run on that environment?\n\n2) we need to submit Pre-trained models by December 4th. What exactly are \"pre-trained models\"?  Does it simply mean December 4th is the last day we can submit stage 1 test data result to check our model performance?\n\nthanks!",
      "votes": null
    },
    {
      "id": "241337",
      "postDate": "11/08/2017 15:42:20",
      "content": "<p>Hi DeltoiX,</p>\n\n<p>The code must be structured in such a way that, provided with appropriate versioning, etc., would allow the sponsors to  reproduce your score.</p>\n\n<p>Pre-trained models bring in already trained, external scripts to kickstart or assist your model development. If you did not use any pre-trained models, then this is a non-issue.</p>\n\n<p>Thanks!</p>",
      "rawMarkdown": "Hi DeltoiX,\n\nThe code must be structured in such a way that, provided with appropriate versioning, etc., would allow the sponsors to  reproduce your score.\n\nPre-trained models bring in already trained, external scripts to kickstart or assist your model development. If you did not use any pre-trained models, then this is a non-issue.\n\nThanks!",
      "votes": null
    },
    {
      "id": "246096",
      "postDate": "11/20/2017 15:05:17",
      "content": "<p>Hi Addison,</p>\n\n<p>thanks for the reply. Few follow up questions:\n1) by \"appropriate versioning\", I suppose you mean using the right version of library?</p>\n\n<p>2) I suppose that the sponsors, in order to reproduce the score, simply run the model code on their computer against stage 2 test data. If so, can I assume the computer to be some kind of very powerful  workstation?  Asking because I am using \"expensive\" cloud environment to execute my model since the model ingests large file format requiring ~10GB of memory at least.</p>\n\n<p>3) So if I use pre-trained model, I simply state what I use in the Official External Data Thread?</p>\n\n<p>please clarify my final set of questions. </p>\n\n<p>thank you!</p>",
      "rawMarkdown": "Hi Addison,\n\nthanks for the reply. Few follow up questions:\n1) by \"appropriate versioning\", I suppose you mean using the right version of library?\n\n2) I suppose that the sponsors, in order to reproduce the score, simply run the model code on their computer against stage 2 test data. If so, can I assume the computer to be some kind of very powerful  workstation?  Asking because I am using \"expensive\" cloud environment to execute my model since the model ingests large file format requiring ~10GB of memory at least.\n\n3) So if I use pre-trained model, I simply state what I use in the Official External Data Thread?\n\nplease clarify my final set of questions. \n\nthank you!",
      "votes": null
    },
    {
      "id": "246176",
      "postDate": "11/20/2017 16:36:55",
      "content": "<p>Hi again,</p>\n\n<p>1) That is correct.\n2) Yes, you can assume that the responsibility of reproduction is on the host.\n3) Correct.</p>",
      "rawMarkdown": "Hi again,\n\n1) That is correct.\n2) Yes, you can assume that the responsibility of reproduction is on the host.\n3) Correct.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 241337,
      "author_name": "addisonhoward",
      "author_url": "",
      "post_date": "11/08/2017 15:42:20",
      "content": "<p>Hi DeltoiX,</p>\n\n<p>The code must be structured in such a way that, provided with appropriate versioning, etc., would allow the sponsors to  reproduce your score.</p>\n\n<p>Pre-trained models bring in already trained, external scripts to kickstart or assist your model development. If you did not use any pre-trained models, then this is a non-issue.</p>\n\n<p>Thanks!</p>",
      "votes": null,
      "replies": [
        {
          "id": 246096,
          "author_name": "deltoix",
          "author_url": "",
          "post_date": "11/20/2017 15:05:17",
          "content": "<p>Hi Addison,</p>\n\n<p>thanks for the reply. Few follow up questions:\n1) by \"appropriate versioning\", I suppose you mean using the right version of library?</p>\n\n<p>2) I suppose that the sponsors, in order to reproduce the score, simply run the model code on their computer against stage 2 test data. If so, can I assume the computer to be some kind of very powerful  workstation?  Asking because I am using \"expensive\" cloud environment to execute my model since the model ingests large file format requiring ~10GB of memory at least.</p>\n\n<p>3) So if I use pre-trained model, I simply state what I use in the Official External Data Thread?</p>\n\n<p>please clarify my final set of questions. </p>\n\n<p>thank you!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 246176,
          "author_name": "addisonhoward",
          "author_url": "",
          "post_date": "11/20/2017 16:36:55",
          "content": "<p>Hi again,</p>\n\n<p>1) That is correct.\n2) Yes, you can assume that the responsibility of reproduction is on the host.\n3) Correct.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "241112": "Hi,\nSorry, this is a newbie question, but\nWe have to submit our model code at the very latest on December 10th. \n\n1) Is there any requirement in terms of how the code should be structured? \n\nI am asking because my understanding that kaggle will verify the validity of our stage 2 label submission by running our code themselves against stage 2 test data. Say my code is written in a way that leverages third party cloud environment. I suppose I need to provide an instruction on how to deploy the code and test data to run on that environment?\n\n2) we need to submit Pre-trained models by December 4th. What exactly are \"pre-trained models\"?  Does it simply mean December 4th is the last day we can submit stage 1 test data result to check our model performance?\n\nthanks!",
    "241337": "Hi DeltoiX,\n\nThe code must be structured in such a way that, provided with appropriate versioning, etc., would allow the sponsors to  reproduce your score.\n\nPre-trained models bring in already trained, external scripts to kickstart or assist your model development. If you did not use any pre-trained models, then this is a non-issue.\n\nThanks!",
    "246096": "Hi Addison,\n\nthanks for the reply. Few follow up questions:\n1) by \"appropriate versioning\", I suppose you mean using the right version of library?\n\n2) I suppose that the sponsors, in order to reproduce the score, simply run the model code on their computer against stage 2 test data. If so, can I assume the computer to be some kind of very powerful  workstation?  Asking because I am using \"expensive\" cloud environment to execute my model since the model ingests large file format requiring ~10GB of memory at least.\n\n3) So if I use pre-trained model, I simply state what I use in the Official External Data Thread?\n\nplease clarify my final set of questions. \n\nthank you!",
    "246176": "Hi again,\n\n1) That is correct.\n2) Yes, you can assume that the responsibility of reproduction is on the host.\n3) Correct."
  },
  "source": "meta"
}