{
  "id": 186830,
  "title": "Can anyone could fix this issue",
  "url": "/competitions/landmark-recognition-2020/discussion/186830",
  "author_name": "Mohammed Rizin V K",
  "post_date": "2020-09-26T08:18:51.172000",
  "votes": 0,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I could train the model on kaggle without any memory issues. but couldnt save the weights of it in the free 5.12 GB Space but on colab i couldnt  train the network but allows to store the weight file in my drive . Can anyone fix this issue.<br>\nI used ef0 + follow some architecture(<br>\none way,<br>\nwas to reduce the batch but when reducing the batch size to run on colab gives bad result from 15 loss to 20 loss(when reducing) and accuracy from 1% to 0.3% on first epoch</p>\n<p>Can anyone fix this issue</p>",
  "messages": [
    {
      "id": 1027538,
      "postDate": "2020-09-26T08:18:51.173Z",
      "content": "<p>I could train the model on kaggle without any memory issues. but couldnt save the weights of it in the free 5.12 GB Space but on colab i couldnt  train the network but allows to store the weight file in my drive . Can anyone fix this issue.<br>\nI used ef0 + follow some architecture(<br>\none way,<br>\nwas to reduce the batch but when reducing the batch size to run on colab gives bad result from 15 loss to 20 loss(when reducing) and accuracy from 1% to 0.3% on first epoch</p>\n<p>Can anyone fix this issue</p>",
      "rawMarkdown": "I could train the model on kaggle without any memory issues. but couldnt save the weights of it in the free 5.12 GB Space but on colab i couldnt  train the network but allows to store the weight file in my drive . Can anyone fix this issue.\nI used ef0 + follow some architecture(\none way,\nwas to reduce the batch but when reducing the batch size to run on colab gives bad result from 15 loss to 20 loss(when reducing) and accuracy from 1% to 0.3% on first epoch\n\nCan anyone fix this issue"
    },
    {
      "id": 1029942,
      "postDate": "2020-09-28T09:37:56.673Z",
      "rawMarkdown": "",
      "votes": -1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 1029942,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-09-28T09:37:56.673000",
      "content": "",
      "votes": -1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1027538": "I could train the model on kaggle without any memory issues. but couldnt save the weights of it in the free 5.12 GB Space but on colab i couldnt  train the network but allows to store the weight file in my drive . Can anyone fix this issue.\nI used ef0 + follow some architecture(\none way,\nwas to reduce the batch but when reducing the batch size to run on colab gives bad result from 15 loss to 20 loss(when reducing) and accuracy from 1% to 0.3% on first epoch\n\nCan anyone fix this issue",
    "1029942": ""
  }
}