{
  "id": 145959,
  "title": "What's unique about tf.data?",
  "url": "/competitions/flower-classification-with-tpus/discussion/145959",
  "author_name": "Kurian Benoy",
  "post_date": "2020-04-25T09:05:50.184000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>One of the most standalone features of tf.data is we can dynamically decide the level of parallelism to use (tf.data.experimental.AUTOTUNE).</p>\n\n<p>This makes TensorFlow totally different from other frameworks. </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1310697%2F0664147ec00d55c27550587aa834b20a%2FScreenshot_2020-04-25%20Google%20MENA%20Digital%20Days%20Deck%20-%20Sayak.png?generation=1587805381551721&amp;alt=media\" alt=\"\"></p>\n\n<p>Check out the excellent talk by <a href=\"/spsayakpaul\">@spsayakpaul</a> here:</p>\n\n<p><a href=\"https://youtu.be/fc0_eLmxm0E\">Talk link</a>\n<a href=\"https://docs.google.com/presentation/d/1j50c2oXttYEUdrhgEs86kI5vKE1bj8O5L4YXLMbMmF4/edit#slide=id.g6fc5c43911_0_20\">Deck link</a></p>",
  "messages": [
    {
      "id": 820234,
      "postDate": "2020-04-25T09:05:50.183Z",
      "content": "<p>One of the most standalone features of tf.data is we can dynamically decide the level of parallelism to use (tf.data.experimental.AUTOTUNE).</p>\n\n<p>This makes TensorFlow totally different from other frameworks. </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1310697%2F0664147ec00d55c27550587aa834b20a%2FScreenshot_2020-04-25%20Google%20MENA%20Digital%20Days%20Deck%20-%20Sayak.png?generation=1587805381551721&amp;alt=media\" alt=\"\"></p>\n\n<p>Check out the excellent talk by <a href=\"/spsayakpaul\">@spsayakpaul</a> here:</p>\n\n<p><a href=\"https://youtu.be/fc0_eLmxm0E\">Talk link</a>\n<a href=\"https://docs.google.com/presentation/d/1j50c2oXttYEUdrhgEs86kI5vKE1bj8O5L4YXLMbMmF4/edit#slide=id.g6fc5c43911_0_20\">Deck link</a></p>",
      "rawMarkdown": "One of the most standalone features of tf.data is we can dynamically decide the level of parallelism to use (tf.data.experimental.AUTOTUNE).\n\nThis makes TensorFlow totally different from other frameworks. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1310697%2F0664147ec00d55c27550587aa834b20a%2FScreenshot_2020-04-25%20Google%20MENA%20Digital%20Days%20Deck%20-%20Sayak.png?generation=1587805381551721&amp;alt=media)\n\nCheck out the excellent talk by @spsayakpaul here:\n\n[Talk link](https://youtu.be/fc0_eLmxm0E)\n[Deck link](https://docs.google.com/presentation/d/1j50c2oXttYEUdrhgEs86kI5vKE1bj8O5L4YXLMbMmF4/edit#slide=id.g6fc5c43911_0_20)\n",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "820234": "One of the most standalone features of tf.data is we can dynamically decide the level of parallelism to use (tf.data.experimental.AUTOTUNE).\n\nThis makes TensorFlow totally different from other frameworks. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1310697%2F0664147ec00d55c27550587aa834b20a%2FScreenshot_2020-04-25%20Google%20MENA%20Digital%20Days%20Deck%20-%20Sayak.png?generation=1587805381551721&amp;alt=media)\n\nCheck out the excellent talk by @spsayakpaul here:\n\n[Talk link](https://youtu.be/fc0_eLmxm0E)\n[Deck link](https://docs.google.com/presentation/d/1j50c2oXttYEUdrhgEs86kI5vKE1bj8O5L4YXLMbMmF4/edit#slide=id.g6fc5c43911_0_20)\n"
  }
}