{
  "id": 430057,
  "title": "What are the best practices for accessing the dataset and running the model? ",
  "url": "/competitions/asl-fingerspelling/discussion/430057",
  "author_name": "MUSTAFA ALSHUALA",
  "post_date": "2023-08-08T06:06:59.131000",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello everyone, <br>\nWhat are the best practices for: <br>\n<strong>Accessing dataset?</strong><br>\nOption 1: Copy dataset from Kaggle to Google Drive? <br>\nOption 2: Accessing dataset by using Kaggle notebook?<br>\nOption 3: ? <br>\nThe issue with the first option is that copy stopped in 90% saying that there is no space left although I subscribed with the 2TB in Google Drive. On the other hand, the issue with the second option is that, I would like to utilize my Google Colab Pro. </p>\n<p>**Running the model **</p>\n<ol>\n<li><p>Google Colab Pro ? <br>\nA100GPU? <br>\nV100GPU?<br>\nT4 GPU?<br>\nTPU?</p></li>\n<li><p>? </p></li>\n</ol>\n<p>Your advice is highly appreciated, </p>\n<p>Best Regards,<br>\nMustafa </p>",
  "messages": [
    {
      "id": 2379384,
      "postDate": "2023-08-08T06:23:19.793Z",
      "content": "<p>Best practice on colab GPU is probably creating a tfrecords dataset on Kaggle and <a href=\"https://www.kaggle.com/discussions/general/156610\" target=\"_blank\">downloading it straight to colab notebook</a>.<br>\nOn colab TPU you will have to use google cloud storage which is a bit more complicated.</p>",
      "rawMarkdown": "Best practice on colab GPU is probably creating a tfrecords dataset on Kaggle and [downloading it straight to colab notebook](https://www.kaggle.com/discussions/general/156610).\nOn colab TPU you will have to use google cloud storage which is a bit more complicated.",
      "votes": 1,
      "replies": [
        {
          "id": 2379396,
          "postDate": "2023-08-08T06:31:00.670Z",
          "content": "<p>Thank you so much for the advice. </p>",
          "rawMarkdown": "Thank you so much for the advice. "
        }
      ]
    },
    {
      "id": 2379347,
      "postDate": "2023-08-08T06:06:59.130Z",
      "content": "<p>Hello everyone, <br>\nWhat are the best practices for: <br>\n<strong>Accessing dataset?</strong><br>\nOption 1: Copy dataset from Kaggle to Google Drive? <br>\nOption 2: Accessing dataset by using Kaggle notebook?<br>\nOption 3: ? <br>\nThe issue with the first option is that copy stopped in 90% saying that there is no space left although I subscribed with the 2TB in Google Drive. On the other hand, the issue with the second option is that, I would like to utilize my Google Colab Pro. </p>\n<p>**Running the model **</p>\n<ol>\n<li><p>Google Colab Pro ? <br>\nA100GPU? <br>\nV100GPU?<br>\nT4 GPU?<br>\nTPU?</p></li>\n<li><p>? </p></li>\n</ol>\n<p>Your advice is highly appreciated, </p>\n<p>Best Regards,<br>\nMustafa </p>",
      "rawMarkdown": "Hello everyone, \nWhat are the best practices for: \n**Accessing dataset?**\nOption 1: Copy dataset from Kaggle to Google Drive? \nOption 2: Accessing dataset by using Kaggle notebook?\nOption 3: ? \nThe issue with the first option is that copy stopped in 90% saying that there is no space left although I subscribed with the 2TB in Google Drive. On the other hand, the issue with the second option is that, I would like to utilize my Google Colab Pro. \n\n**Running the model **\n1. Google Colab Pro ? \nA100GPU? \nV100GPU?\nT4 GPU?\nTPU?\n\n2. ? \n\nYour advice is highly appreciated, \n\nBest Regards,\nMustafa ",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 2379384,
      "author_name": "greySnow",
      "author_url": "",
      "post_date": "2023-08-08T06:23:19.793000",
      "content": "<p>Best practice on colab GPU is probably creating a tfrecords dataset on Kaggle and <a href=\"https://www.kaggle.com/discussions/general/156610\" target=\"_blank\">downloading it straight to colab notebook</a>.<br>\nOn colab TPU you will have to use google cloud storage which is a bit more complicated.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 2379396,
          "author_name": "MUSTAFA ALSHUALA",
          "author_url": "",
          "post_date": "2023-08-08T06:31:00.670000",
          "content": "<p>Thank you so much for the advice. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2379384": "Best practice on colab GPU is probably creating a tfrecords dataset on Kaggle and [downloading it straight to colab notebook](https://www.kaggle.com/discussions/general/156610).\nOn colab TPU you will have to use google cloud storage which is a bit more complicated.",
    "2379347": "Hello everyone, \nWhat are the best practices for: \n**Accessing dataset?**\nOption 1: Copy dataset from Kaggle to Google Drive? \nOption 2: Accessing dataset by using Kaggle notebook?\nOption 3: ? \nThe issue with the first option is that copy stopped in 90% saying that there is no space left although I subscribed with the 2TB in Google Drive. On the other hand, the issue with the second option is that, I would like to utilize my Google Colab Pro. \n\n**Running the model **\n1. Google Colab Pro ? \nA100GPU? \nV100GPU?\nT4 GPU?\nTPU?\n\n2. ? \n\nYour advice is highly appreciated, \n\nBest Regards,\nMustafa "
  }
}