{
  "id": 254924,
  "title": "Modelling And Initailized Random Weights",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/254924",
  "author_name": "Siaw-Darko Emmanuel",
  "post_date": "2021-07-24T12:17:01.490000",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hi I'm somewhat new to kaggle competitions and I'd like to know how to ensure reproducibility when running with deep learning frameworks like pytorch, tensorflow, pytorch lightning and fastai, unfortunately every time i rerun a notebook different results are generated and i guess it has to do with the randomly initialized weights. <br>\nHow do i prevent this from happening ?</p>",
  "messages": [
    {
      "id": 1398706,
      "postDate": "2021-07-24T12:17:01.490Z",
      "content": "<p>Hi I'm somewhat new to kaggle competitions and I'd like to know how to ensure reproducibility when running with deep learning frameworks like pytorch, tensorflow, pytorch lightning and fastai, unfortunately every time i rerun a notebook different results are generated and i guess it has to do with the randomly initialized weights. <br>\nHow do i prevent this from happening ?</p>",
      "rawMarkdown": "Hi I'm somewhat new to kaggle competitions and I'd like to know how to ensure reproducibility when running with deep learning frameworks like pytorch, tensorflow, pytorch lightning and fastai, unfortunately every time i rerun a notebook different results are generated and i guess it has to do with the randomly initialized weights. \nHow do i prevent this from happening ?",
      "votes": 1
    },
    {
      "id": 1400389,
      "postDate": "2021-07-26T08:50:57.517Z",
      "content": "<p><a href=\"https://www.kaggle.com/weka511\" target=\"_blank\">@weka511</a>  I tried that when i modelled the problem with tensorflow and pytorch.. but i got the same issue… currently trying out other frameworks to see if i can make it work</p>",
      "rawMarkdown": "@weka511  I tried that when i modelled the problem with tensorflow and pytorch.. but i got the same issue... currently trying out other frameworks to see if i can make it work",
      "replies": [
        {
          "id": 1400977,
          "postDate": "2021-07-26T19:10:57.357Z",
          "rawMarkdown": "",
          "isDeleted": true
        },
        {
          "id": 1401148,
          "postDate": "2021-07-27T03:47:47.123Z",
          "content": "<p>thanks <a href=\"https://www.kaggle.com/weka511\" target=\"_blank\">@weka511</a> .. it works now</p>",
          "rawMarkdown": "thanks @weka511 .. it works now"
        }
      ]
    },
    {
      "id": 1400346,
      "postDate": "2021-07-26T08:17:08.813Z",
      "rawMarkdown": "",
      "votes": 1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 1400389,
      "author_name": "Siaw-Darko Emmanuel",
      "author_url": "",
      "post_date": "2021-07-26T08:50:57.517000",
      "content": "<p><a href=\"https://www.kaggle.com/weka511\" target=\"_blank\">@weka511</a>  I tried that when i modelled the problem with tensorflow and pytorch.. but i got the same issue… currently trying out other frameworks to see if i can make it work</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1400977,
          "author_name": "",
          "author_url": "",
          "post_date": "2021-07-26T19:10:57.357000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1401148,
          "author_name": "Siaw-Darko Emmanuel",
          "author_url": "",
          "post_date": "2021-07-27T03:47:47.123000",
          "content": "<p>thanks <a href=\"https://www.kaggle.com/weka511\" target=\"_blank\">@weka511</a> .. it works now</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1400346,
      "author_name": "",
      "author_url": "",
      "post_date": "2021-07-26T08:17:08.813000",
      "content": "",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1398706": "Hi I'm somewhat new to kaggle competitions and I'd like to know how to ensure reproducibility when running with deep learning frameworks like pytorch, tensorflow, pytorch lightning and fastai, unfortunately every time i rerun a notebook different results are generated and i guess it has to do with the randomly initialized weights. \nHow do i prevent this from happening ?",
    "1400389": "@weka511  I tried that when i modelled the problem with tensorflow and pytorch.. but i got the same issue... currently trying out other frameworks to see if i can make it work",
    "1400346": ""
  }
}