Fix typos in migration_debugging.ipynb

Made changes in the lines 103,127,589,625.

Author: synandi

site/en/guide/migrate/migration_debugging.ipynb

@@ -100,7 +100,7 @@
         "        disabled\n",
         "    *   Align random number generation, check numerical equivalence in inference\n",
         "    *   (Optional) Check checkpoints are loaded properly and TF1.x/TF2 models\n",
-        "        generate identitcal output\n",
+        "        generate identical output\n",
         "\n",
         "    b. On single GPU/TPU device\n",
         "\n",
@@ -124,7 +124,7 @@
         "\n",
         "    d. With multi-device strategies (check the intro for [MultiProcessRunner](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/distribute/multi_process_runner.py#L108) at the bottom)\n",
         "\n",
-        "4.  End-to-end covergence testing on real dataset\n",
+        "4.  End-to-end convergence testing on real dataset\n",
         "\n",
         "    a. Check training behaviors with TensorBoard\n",
         "\n",
@@ -586,7 +586,7 @@
         "    # adopt different tolerance strategies before and after 10 steps\n",
         "    first_n_step = 10\n",
         "\n",
-        "    # abosolute difference is limited below 1e-5\n",
+        "    # absolute difference is limited below 1e-5\n",
         "    # set `equal_nan` to be False to detect potential NaN loss issues\n",
         "    abosolute_tolerance = 1e-5\n",
         "    np.testing.assert_allclose(\n",
@@ -622,7 +622,7 @@
         "\n",
         "tf.print vs print/logging.info\n",
         "\n",
-        "- With configurable arguments, `tf.print` can recursively display show first and last few elements of each dimension for printed tensors. Check the [API docs](https://www.tensorflow.org/api_docs/python/tf/print) for details.\n",
+        "- With configurable arguments, `tf.print` can recursively display the first and last few elements of each dimension for printed tensors. Check the [API docs](https://www.tensorflow.org/api_docs/python/tf/print) for details.\n",
         "- For eager execution, both `print` and `tf.print` print the value of the tensor. But `print` may involve device-to-host copy, which can potentially slow down your code. \n",
         "- For graph mode including usage inside `tf.function`, you need to use `tf.print` to print the actual tensor value. `tf.print` is compiled into an op in the graph, whereas `print` and `logging.info` only log at tracing time, which is often not what you want. \n",
         "- `tf.print` also supports printing composite tensors like `tf.RaggedTensor` and `tf.sparse.SparseTensor`.\n",