fixed intro

Author: kashif

7-estimators-multi-gpus.ipynb

@@ -8,7 +8,9 @@
     "\n",
     "We can train on multiple GPUs directly via `tf.keras`'s distributed strategy scope.\n",
     "\n",
-    "TensorFlow's [Estimators](https://www.tensorflow.org/programmers_guide/estimators) API is another useful way to training models in a distributed environment such as on nodes with multiple GPUs or on many nodes with GPUs. This is particularly useful when training on huge datasets especially when used with the `tf.keras` API. Here we will first present the API for the tiny Fashion-MNIST dataset and then show a practical usecase in the end.\n",
+    "TensorFlow's [Estimators](https://www.tensorflow.org/programmers_guide/estimators) API is another useful way to training models in a distributed environment such as on nodes with multiple GPUs or on many nodes with GPUs. This is particularly useful when training on huge datasets especially when used with the `tf.keras` API. \n",
+    "\n",
+    "Here we will first present the `tf.keras` API for the tiny Fashion-MNIST dataset and then show a practical usecase in the end via Estimators.\n",
     "\n",
     "**TL;DR**: Essentially what we want to remember is that a `tf.keras.Model` can be trained with `tf.estimator` API by converting it to an `tf.estimator.Estimator` object via the `tf.keras.estimator.model_to_estimator` method. Once converted we can apply the machinery that `Estimator` provides to train on different hardware configurations."
    ]