{
  "id": 130017,
  "title": "Keras in TPU?",
  "url": "/competitions/flower-classification-with-tpus/discussion/130017",
  "author_name": "",
  "post_date": "2020-02-11T17:06:35.397240700Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi I don't know the basics of tensorflow however I use Keras with tensorflow backend as for deep learning.\nI just want to know whether keras is supported by tpu.\nAlso having fairly good understanding of keras, how difficult it would be to learn tensorflow. </p>",
  "messages": [
    {
      "id": "742963",
      "postDate": "02/11/2020 17:06:35",
      "content": "<p>Hi I don't know the basics of tensorflow however I use Keras with tensorflow backend as for deep learning.\nI just want to know whether keras is supported by tpu.\nAlso having fairly good understanding of keras, how difficult it would be to learn tensorflow. </p>",
      "rawMarkdown": "Hi I don't know the basics of tensorflow however I use Keras with tensorflow backend as for deep learning.\nI just want to know whether keras is supported by tpu.\nAlso having fairly good understanding of keras, how difficult it would be to learn tensorflow.",
      "votes": null
    },
    {
      "id": "743073",
      "postDate": "02/11/2020 19:02:17",
      "content": "<p>Yes, TPUs are supported in tf.keras as of Tensorflow 2.1\nIf you already know Keras, the only necessary addition to your knowledge is the tf.data.Dataset API. Keras is today the high-level API of Tensorflow.\nStart with the <a href=\"https://www.kaggle.com/docs/tpu\">TPU one-pager</a> on Kaggle.\nIf you want to dive deeper, you can go through this code lab: <a href=\"https://codelabs.developers.google.com/codelabs/keras-flowers-tpu/#0\">Keras and modern convnets, on TPUs</a>. It teaches everything you will need, including the basics of tf.keras.</p>",
      "rawMarkdown": "Yes, TPUs are supported in tf.keras as of Tensorflow 2.1\nIf you already know Keras, the only necessary addition to your knowledge is the tf.data.Dataset API. Keras is today the high-level API of Tensorflow.\nStart with the [TPU one-pager](https://www.kaggle.com/docs/tpu) on Kaggle.\nIf you want to dive deeper, you can go through this code lab: [Keras and modern convnets, on TPUs](https://codelabs.developers.google.com/codelabs/keras-flowers-tpu/#0). It teaches everything you will need, including the basics of tf.keras.",
      "votes": null
    },
    {
      "id": "743078",
      "postDate": "02/11/2020 19:06:34",
      "content": "<p>Thank you Martin</p>",
      "rawMarkdown": "Thank you Martin",
      "votes": null
    },
    {
      "id": "744239",
      "postDate": "02/12/2020 17:15:40",
      "content": "<p>In TensorFlow 2.0+, Keras is incorporated into TensorFlow. For example in TF 1.14, we need to <code>import keras</code> and write stuff like <code>'keras.layers.Dense()</code> but in TF 2.0+, we just <code>import tensorflow as tf</code> and write <code>tf.keras.layers.Dense()</code>.</p>",
      "rawMarkdown": "In TensorFlow 2.0+, Keras is incorporated into TensorFlow. For example in TF 1.14, we need to `import keras` and write stuff like `'keras.layers.Dense()` but in TF 2.0+, we just `import tensorflow as tf` and write `tf.keras.layers.Dense()`.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 743073,
      "author_name": "mgorner",
      "author_url": "",
      "post_date": "02/11/2020 19:02:17",
      "content": "<p>Yes, TPUs are supported in tf.keras as of Tensorflow 2.1\nIf you already know Keras, the only necessary addition to your knowledge is the tf.data.Dataset API. Keras is today the high-level API of Tensorflow.\nStart with the <a href=\"https://www.kaggle.com/docs/tpu\">TPU one-pager</a> on Kaggle.\nIf you want to dive deeper, you can go through this code lab: <a href=\"https://codelabs.developers.google.com/codelabs/keras-flowers-tpu/#0\">Keras and modern convnets, on TPUs</a>. It teaches everything you will need, including the basics of tf.keras.</p>",
      "votes": null,
      "replies": [
        {
          "id": 743078,
          "author_name": "abhijithchandradas",
          "author_url": "",
          "post_date": "02/11/2020 19:06:34",
          "content": "<p>Thank you Martin</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 744239,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "02/12/2020 17:15:40",
      "content": "<p>In TensorFlow 2.0+, Keras is incorporated into TensorFlow. For example in TF 1.14, we need to <code>import keras</code> and write stuff like <code>'keras.layers.Dense()</code> but in TF 2.0+, we just <code>import tensorflow as tf</code> and write <code>tf.keras.layers.Dense()</code>.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "742963": "Hi I don't know the basics of tensorflow however I use Keras with tensorflow backend as for deep learning.\nI just want to know whether keras is supported by tpu.\nAlso having fairly good understanding of keras, how difficult it would be to learn tensorflow.",
    "743073": "Yes, TPUs are supported in tf.keras as of Tensorflow 2.1\nIf you already know Keras, the only necessary addition to your knowledge is the tf.data.Dataset API. Keras is today the high-level API of Tensorflow.\nStart with the [TPU one-pager](https://www.kaggle.com/docs/tpu) on Kaggle.\nIf you want to dive deeper, you can go through this code lab: [Keras and modern convnets, on TPUs](https://codelabs.developers.google.com/codelabs/keras-flowers-tpu/#0). It teaches everything you will need, including the basics of tf.keras.",
    "743078": "Thank you Martin",
    "744239": "In TensorFlow 2.0+, Keras is incorporated into TensorFlow. For example in TF 1.14, we need to `import keras` and write stuff like `'keras.layers.Dense()` but in TF 2.0+, we just `import tensorflow as tf` and write `tf.keras.layers.Dense()`."
  },
  "source": "meta"
}