Replace tf.{data.experimental,contrib.data}.unbatch() with Dataset.unbatch().

PiperOrigin-RevId: 273403022

Replace tf.{data.experimental,contrib.data}.unbatch() with Dataset.unbatch().
a5b3c6b5 · Derek Murray · tensorflow-copybara · 1620b6d3 · a5b3c6b5
Commit a5b3c6b5 authored 5 years ago by Derek Murray Committed by tensorflow-copybara 5 years ago
--- a/docs/tutorials/federated_learning_for_text_generation.ipynb
+++ b/docs/tutorials/federated_learning_for_text_generation.ipynb
@@ -468,7 +468,7 @@
        "      # Map ASCII chars to int64 indexes using the vocab\n",
        "      dataset.map(to_ids)\n",
        "      # Split into individual chars\n",
-        "      .apply(tf.data.experimental.unbatch())\n",
+        "      .unbatch()\n",
        "      # Form example sequences of SEQ_LENGTH +1\n",
        "      .batch(SEQ_LENGTH + 1, drop_remainder=True)\n",
        "      # Shuffle and form minibatches\n",

 %% Cell type:markdown id: tags:

 ##### Copyright 2019 The TensorFlow Authors.

 %% Cell type:code id: tags:

 ``` 
 #@title Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at
 #
 # https://www.apache.org/licenses/LICENSE-2.0
 #
 # Unless required by applicable law or agreed to in writing, software
 # distributed under the License is distributed on an "AS IS" BASIS,
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
 ```

 %% Cell type:markdown id: tags:

 # Federated Learning for Text Generation

 %% Cell type:markdown id: tags:

 <table class="tfo-notebook-buttons" align="left">
  <td>
    <a target="_blank" href="https://www.tensorflow.org/federated/tutorials/federated_learning_for_text_generation"><img src="https://www.tensorflow.org/images/tf_logo_32px.png" />View on TensorFlow.org</a>
  </td>
  <td>
    <a target="_blank" href="https://colab.research.google.com/github/tensorflow/federated/blob/v0.9.0/docs/tutorials/federated_learning_for_text_generation.ipynb"><img src="https://www.tensorflow.org/images/colab_logo_32px.png" />Run in Google Colab</a>
  </td>
  <td>
    <a target="_blank" href="https://github.com/tensorflow/federated/blob/v0.9.0/docs/tutorials/federated_learning_for_text_generation.ipynb"><img src="https://www.tensorflow.org/images/GitHub-Mark-32px.png" />View source on GitHub</a>
  </td>
 </table>

 %% Cell type:markdown id: tags:

 **NOTE**: This colab has been verified to work with the [latest released version](https://github.com/tensorflow/federated#compatibility) of the `tensorflow_federated` pip package, but the Tensorflow Federated project is still in pre-release development and may not work on `master`.

 This tutorial builds on the concepts in the [Federated Learning for Image Classification](federated_learning_for_image_classification.md) tutorial, and demonstrates several other useful approaches for federated learning.

 In particular, we load a previously trained Keras model, and refine it using  federated training on a (simulated) decentralized dataset. This is practically important for several reasons . The ability to use serialized models makes it easy to mix federated learning with other ML approaches. Further, this allows use of an increasing range of pre-trained models --- for example, training language models from scratch is rarely necessary, as numerous pre-trained models are now widely available (see, e.g., [TF Hub](https://www.tensorflow.org/hub)). Instead, it makes more sense to start from a pre-trained model, and refine it using Federated Learning, adapting to the particular characteristics of the decentralized data for a particular application.

 For this tutorial, we start with a RNN that generates ASCII characters, and refine it via federated learning. We also show how the final weights can be fed back to the original Keras model, allowing easy evaluation and text generation using standard tools.

 %% Cell type:code id: tags:

 ``` 
 # NOTE: If you are running a Jupyter notebook, and installing a locally built
 # pip package, you may need to edit the following to point to the '.whl' file
 # on your local filesystem.

 !pip install --quiet tensorflow_federated
 !pip install --quiet tf-nightly

 # NOTE: Jupyter requires a patch to asyncio.
 !pip install --upgrade nest_asyncio
 import nest_asyncio
 nest_asyncio.apply()
 ```

 %% Cell type:code id: tags:

 ``` 
 from __future__ import absolute_import, division, print_function

 import collections
 import functools
 import os
 import six
 import time

 import numpy as np
 import tensorflow as tf
 import tensorflow_federated as tff

 tf.compat.v1.enable_v2_behavior()

 np.random.seed(0)

 # Test the TFF is working:
 tff.federated_computation(lambda: 'Hello, World!')()
 ```

 %% Output

    'Hello, World!'

 %% Cell type:markdown id: tags:

 # Load a pre-trained model

 We load a model that was pre-trained following the TensorFlow tutorial
 [Text generation using a RNN with eager execution](https://www.tensorflow.org/tutorials/sequences/text_generation). However,
 rather than training on [The Complete Works of Shakespeare](http://www.gutenberg.org/files/100/100-0.txt), we pre-trained the model on the text from the Charles Dickens'
    [A Tale of Two Cities](http://www.ibiblio.org/pub/docs/books/gutenberg/9/98/98.txt)
    and
    [A Christmas Carol](http://www.ibiblio.org/pub/docs/books/gutenberg/4/46/46.txt).

 Other than expanding the vocabularly, we didn't modify the original tutorial, so this initial model isn't state-of-the-art, but it produces reasonable predictions and is sufficient for our tutorial purposes. The final model was saved with `tf.keras.models.save_model(include_optimizer=False)`.

 We will use federated learning to fine-tune this model for Shakespeare in this tutorial, using a federated version of the data provided by TFF.


 %% Cell type:markdown id: tags:

 ## Generate the vocab lookup tables

 %% Cell type:code id: tags:

 ``` 
 # A fixed vocabularly of ASCII chars that occur in the works of Shakespeare and Dickens:
 vocab = list('dhlptx@DHLPTX $(,048cgkoswCGKOSW[_#\'/37;?bfjnrvzBFJNRVZ"&*.26:\naeimquyAEIMQUY]!%)-159\r')

 # Creating a mapping from unique characters to indices
 char2idx = {u:i for i, u in enumerate(vocab)}
 idx2char = np.array(vocab)
 ```

 %% Cell type:markdown id: tags:

 ## Load the pre-trained model and generate some text

 %% Cell type:code id: tags:

 ``` 
 def load_model(batch_size):
  urls = {
      1: 'https://storage.googleapis.com/tff-models-public/dickens_rnn.batch1.kerasmodel',
      8: 'https://storage.googleapis.com/tff-models-public/dickens_rnn.batch8.kerasmodel'}
  assert batch_size in urls, 'batch_size must be in ' + str(urls.keys())
  url = urls[batch_size]
  local_file = tf.keras.utils.get_file(os.path.basename(url), origin=url)
  return tf.keras.models.load_model(local_file, compile=False)
 ```

 %% Cell type:code id: tags:

 ``` 
 def generate_text(model, start_string):
  # From https://www.tensorflow.org/tutorials/sequences/text_generation
  num_generate = 200
  input_eval = [char2idx[s] for s in start_string]
  input_eval = tf.expand_dims(input_eval, 0)
  text_generated = []
  temperature = 1.0

  model.reset_states()
  for i in range(num_generate):
    predictions = model(input_eval)
    predictions = tf.squeeze(predictions, 0)
    predictions = predictions / temperature
    predicted_id = tf.random.categorical(
        predictions, num_samples=1)[-1, 0].numpy()
    input_eval = tf.expand_dims([predicted_id], 0)
    text_generated.append(idx2char[predicted_id])

  return (start_string + ''.join(text_generated))
 ```

 %% Cell type:code id: tags:

 ``` 
 # Text generation requires a batch_size=1 model.
 keras_model_batch1 = load_model(batch_size=1)
 print(generate_text(keras_model_batch1, 'What of TensorFlow Federated, you ask? '))
 ```

 %% Output

    Downloading data from https://storage.googleapis.com/tff-models-public/dickens_rnn.batch1.kerasmodel
    16195584/16193984 [==============================] - 0s 0us/step
    16203776/16193984 [==============================] - 0s 0us/step
    What of TensorFlow Federated, you ask? Stryver, seemed, unaternight,
    Fruncied eyebrows at his forgery and the rest of its contempt.
    
    Mr. Cruncher had made his opportunity outsidere Stryver, even had some
    fardens--natured impossible wor

 %% Cell type:markdown id: tags:

 # Load and Preprocess the Federated Shakespeare Data

 The `tff.simulation.datasets` package provides a variety of datasets that are split into "clients", where each client corresponds to a dataset on a particular device that might participate in federated learning.

 These datasets provide realistic non-IID data distributions that replicate in simulation the challenges of training on real decentralized data. Some of the pre-processing of this data was done using tools from the [Leaf project](https://arxiv.org/abs/1812.01097) ([github](https://github.com/TalwalkarLab/leaf)).

 %% Cell type:code id: tags:

 ``` 
 train_data, test_data = tff.simulation.datasets.shakespeare.load_data()
 ```

 %% Cell type:markdown id: tags:

 The datasets provided by `shakespeare.load_data()` consist of a sequence of
 string `Tensors`, one for each line spoken by a particular character in a
 Shakespeare play. The client keys consist of the name of the play joined with
 the name of the character, so for example `MUCH_ADO_ABOUT_NOTHING_OTHELLO` corresponds to the lines for the character Othello in the play *Much Ado About Nothing*. Note that in a real federated learning scenario
 clients are never identified or tracked by ids, but for simulation it is useful
 to work with keyed datasets.

 Here, for example, we can look at some data from King Lear:

 %% Cell type:code id: tags:

 ``` 
 # Here the play is "The Tragedy of King Lear" and the character is "King".
 raw_example_dataset = train_data.create_tf_dataset_for_client(
    'THE_TRAGEDY_OF_KING_LEAR_KING')
 # To allow for future extensions, each entry x
 # is an OrderedDict with a single key 'snippets' which contains the text.
 for x in raw_example_dataset.take(2):
  print(x['snippets'])
 ```

 %% Output

    tf.Tensor(, shape=(), dtype=string)
    tf.Tensor(What?, shape=(), dtype=string)

 %% Cell type:markdown id: tags:

 We now use `tf.data.Dataset` transformations to prepare this data for training the char RNN loaded above.

 %% Cell type:code id: tags:

 ``` 
 # Input pre-processing parameters
 SEQ_LENGTH = 100
 BATCH_SIZE = 8
 BUFFER_SIZE = 10000  # For dataset shuffling
 ```

 %% Cell type:code id: tags:

 ``` 
 # Using a namedtuple with keys x and y as the output type of the
 # dataset keeps both TFF and Keras happy:
 BatchType = collections.namedtuple('BatchType', ['x', 'y'])

 # Construct a lookup table to map string chars to indexes,
 # using the vocab loaded above:
 table = tf.lookup.StaticHashTable(
    tf.lookup.KeyValueTensorInitializer(
        keys=vocab, values=tf.constant(list(range(len(vocab))),
                                       dtype=tf.int64)),
    default_value=0)


 def to_ids(x):
  s = tf.reshape(x['snippets'], shape=[1])
  chars = tf.strings.bytes_split(s).values
  ids = table.lookup(chars)
  return ids


 def split_input_target(chunk):
  input_text = tf.map_fn(lambda x: x[:-1], chunk)
  target_text = tf.map_fn(lambda x: x[1:], chunk)
  return BatchType(input_text, target_text)


 def preprocess(dataset):
  return (
      # Map ASCII chars to int64 indexes using the vocab
      dataset.map(to_ids)
      # Split into individual chars
-      .apply(tf.data.experimental.unbatch())
+      .unbatch()
      # Form example sequences of SEQ_LENGTH +1
      .batch(SEQ_LENGTH + 1, drop_remainder=True)
      # Shuffle and form minibatches
      .shuffle(BUFFER_SIZE).batch(BATCH_SIZE, drop_remainder=True)
      # And finally split into (input, target) tuples,
      # each of length SEQ_LENGTH.
      .map(split_input_target))
 ```

 %% Cell type:markdown id: tags:

 Note that in the formation of the original sequences and in the formation of
 batches above, we use `drop_remainder=True` for simplicity. This means that any
 characters (clients) that don't have at least `(SEQ_LENGTH + 1) * BATCH_SIZE`
 chars of text will have empty datasets. A typical approach to address this would
 be to pad the batches with a special token, and then mask the loss to not take
 the padding tokens into account.

 This would complicate the example somewhat, so for this tutorial we only use full batches, as in the
 [standard tutorial](https://www.tensorflow.org/tutorials/sequences/text_generation).
 However, in the federated setting this issue is more significant, because many
 users might have small datasets.

 Now we can preprocess our `raw_example_dataset`, and check the types:

 %% Cell type:code id: tags:

 ``` 
 example_dataset = preprocess(raw_example_dataset)
 print(tf.data.experimental.get_structure(example_dataset))
 ```

 %% Output

    BatchType(x=TensorSpec(shape=(8, 100), dtype=tf.int64, name=None), y=TensorSpec(shape=(8, 100), dtype=tf.int64, name=None))

 %% Cell type:markdown id: tags:

 # Compile the model and test on the preprocessed data

 %% Cell type:markdown id: tags:

 We loaded an uncompiled keras model, but in order to run `keras_model.evaluate`, we need to compile it with a loss and metrics. We will also compile in an optimizer, which will be used as the on-device optimizer in Federated Learning.

 %% Cell type:markdown id: tags:

 The original tutorial didn't have char-level accuracy (the fraction
 of predictions where the highest probability was put on the correct
 next char). This is a useful metric, so we add it.
 However, we need to define a new metric class for this because
 our predictions have rank 3 (a vector of logits for each of the
 `BATCH_SIZE * SEQ_LENGTH` predictions), and `SparseCategoricalAccuracy`
 expects only rank 2 predictions.

 %% Cell type:code id: tags:

 ``` 
 class FlattenedCategoricalAccuracy(tf.keras.metrics.SparseCategoricalAccuracy):

  def __init__(self, name='accuracy', dtype=None):
    super(FlattenedCategoricalAccuracy, self).__init__(name, dtype=dtype)

  def update_state(self, y_true, y_pred, sample_weight=None):
    y_true = tf.reshape(y_true, [-1, 1])
    y_pred = tf.reshape(y_pred, [-1, len(vocab), 1])
    return super(FlattenedCategoricalAccuracy, self).update_state(
        y_true, y_pred, sample_weight)
 ```

 %% Cell type:code id: tags:

 ``` 
 def compile(keras_model):
  keras_model.compile(
      optimizer=tf.keras.optimizers.SGD(lr=0.5),
      loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
      metrics=[FlattenedCategoricalAccuracy()])
  return keras_model
 ```

 %% Cell type:markdown id: tags:

 Now we can compile a model, and evaluate it on our `example_dataset`.

 %% Cell type:code id: tags:

 ``` 
 BATCH_SIZE = 8  # The training and eval batch size for the rest of this tutorial.
 keras_model = load_model(batch_size=BATCH_SIZE)

 compile(keras_model)

 # Confirm that loss is much lower on Shakespeare than on random data
 print('Evaluating on an example Shakespeare character:')
 keras_model.evaluate(example_dataset.take(1))

 # As a sanity check, we can construct some completely random data, where we expect
 # the accuracy to be essentially random:
 random_indexes = np.random.randint(
    low=0, high=len(vocab), size=1 * BATCH_SIZE * (SEQ_LENGTH + 1))
 data = {
    'snippets':
        tf.constant(''.join(np.array(vocab)[random_indexes]), shape=[1, 1])
 }
 random_dataset = preprocess(tf.data.Dataset.from_tensor_slices(data))
 print('\nExpected accuracy for random guessing: {:.3f}'.format(1.0 / len(vocab)))
 print('Evaluating on completely random data:')
 keras_model.evaluate(random_dataset, steps=1)
 ```

 %% Output

    Evaluating on an example Shakespeare character:
          1/Unknown - 1s 1s/step - loss: 2.9296 - accuracy: 0.4412
    Expected accuracy for random guessing: 0.012
    Evaluating on completely random data:
    1/1 [==============================] - 0s 426ms/step - loss: 11.4698 - accuracy: 0.0125

    [11.4698486328125, 0.0125]

 %% Cell type:markdown id: tags:

 # Fine-tune the model with Federated Learning

 %% Cell type:markdown id: tags:

 TFF serializes all TensorFlow computations so they can potentially be run in a
 non-Python environment (even though at the moment, only a simulation runtime implemented in Python is available). Even though we are running in eager mode, (TF 2.0), currently TFF serializes TensorFlow computations by constructing the
 necessary ops inside the context of a "`with tf.Graph.as_default()`" statement.
 Thus, we need to provide a function that TFF can use to introduce our model into
 a graph it controls. We do this as follows:

 %% Cell type:code id: tags:

 ``` 
 # Clone the keras_model inside `create_tff_model()`, which TFF will
 # call to produce a new copy of the model inside the graph that it will serialize.
 def create_tff_model():
  # TFF uses a `dummy_batch` so it knows the types and shapes
  # that your model expects.
  x = tf.constant(np.random.randint(1, len(vocab), size=[BATCH_SIZE, SEQ_LENGTH]))
  dummy_batch = collections.OrderedDict([('x', x), ('y', x)])
  keras_model_clone = compile(tf.keras.models.clone_model(keras_model))
  return tff.learning.from_compiled_keras_model(
      keras_model_clone, dummy_batch=dummy_batch)
 ```

 %% Cell type:markdown id: tags:

 Now we are ready to construct a Federated Averaging iterative process, which we will use to improve the model (for details on the Federated Averaging algorithm, see the paper [Communication-Efficient Learning of Deep Networks from Decentralized Data](https://arxiv.org/abs/1602.05629)).

 We use a compiled Keras model to perform standard (non-federated) evaluation after each round of federated training. This is useful for research purposes when doing simulated federated learning and there is a  standard test dataset.

 In a realistic production setting this same technique might be used to take models trained with federated learning and evaluate them on a centralized benchmark dataset for testing or quality assurance purposes.

 %% Cell type:code id: tags:

 ``` 
 # This command builds all the TensorFlow graphs and serializes them:
 fed_avg = tff.learning.build_federated_averaging_process(model_fn=create_tff_model)
 ```

 %% Cell type:markdown id: tags:

 Here is the simplest possible loop, where we run federated averaging for one round on a single client on a single batch:

 %% Cell type:code id: tags:

 ``` 
 state = fed_avg.initialize()
 state, metrics = fed_avg.next(state, [example_dataset.take(1)])
 print(metrics)
 ```

 %% Output

    <accuracy=0.016249999404,loss=4.45304775238>

 %% Cell type:markdown id: tags:

 Now let's write a slightly more interesting training and evaluation loop.

 So that this simulation still runs relatively quickly,  we train on the same three clients each round, only considering two minibatches for each.


 %% Cell type:code id: tags:

 ``` 
 def data(client, source=train_data):
  return preprocess(
      source.create_tf_dataset_for_client(client)).take(2)

 clients = ['ALL_S_WELL_THAT_ENDS_WELL_CELIA',
           'MUCH_ADO_ABOUT_NOTHING_OTHELLO',
           'THE_TRAGEDY_OF_KING_LEAR_KING']

 train_datasets = [data(client) for client in clients]

 # We concatenate the test datasets for evaluation with Keras.
 test_dataset = functools.reduce(
    lambda d1, d2: d1.concatenate(d2),
    [data(client, test_data) for client in clients])

 # NOTE: If the statement below fails, it means that you are
 # using an older version of TFF without the high-performance
 # executor stack. Call `tff.framework.set_default_executor()`
 # instead to use the default reference runtime.
 if six.PY3:
  tff.framework.set_default_executor(tff.framework.create_local_executor())
 ```

 %% Cell type:markdown id: tags:

 The initial state of the model produced by `fed_avg.initialize()` is based
 on the random initializers for the Keras model, not the weights that were loaded,
 since `clone_model()` does not clone the weights. To start training
 from a pre-trained model, we set the model weights in the server state
 directly from the loaded model.

 %% Cell type:code id: tags:

 ``` 
 NUM_ROUNDS = 3

 # The state of the FL server, containing the model and optimization state.
 state = fed_avg.initialize()

 state = tff.learning.state_with_new_model_weights(
    state,
    trainable_weights=[v.numpy() for v in keras_model.trainable_weights],
    non_trainable_weights=[
        v.numpy() for v in keras_model.non_trainable_weights
    ])


 def keras_evaluate(state, round_num):
  tff.learning.assign_weights_to_keras_model(keras_model, state.model)
  print('Evaluating before training round', round_num)
  keras_model.evaluate(example_dataset, steps=2)


 for round_num in range(NUM_ROUNDS):
  keras_evaluate(state, round_num)
  # N.B. The TFF runtime is currently fairly slow,
  # expect this to get significantly faster in future releases.
  state, metrics = fed_avg.next(state, train_datasets)
  print('Training metrics: ', metrics)

 keras_evaluate(state, NUM_ROUNDS + 1)
 ```

 %% Output

    Evaluating before training round 0
    2/2 [==============================] - 1s 649ms/step - loss: 3.2126 - accuracy: 0.4288
    Training metrics:  <accuracy=0.413541674614,loss=3.30218100548>
    Evaluating before training round 1
    2/2 [==============================] - 1s 561ms/step - loss: 2.8153 - accuracy: 0.4588
    Training metrics:  <accuracy=0.434791654348,loss=2.98462796211>
    Evaluating before training round 2
    2/2 [==============================] - 1s 601ms/step - loss: 2.7724 - accuracy: 0.4575
    Training metrics:  <accuracy=0.437916666269,loss=2.87845301628>
    Evaluating before training round 4
    2/2 [==============================] - 1s 645ms/step - loss: 2.7769 - accuracy: 0.4444

 %% Cell type:markdown id: tags:

 With the default changes, we haven't done enough training to make a big difference, but if you train longer on more Shakespeare data, you should see a difference in the style of the text generated with the updated model:

 %% Cell type:code id: tags:

 ``` 
 keras_model_batch1.set_weights([v.numpy() for v in keras_model.weights])
 # Text generation requires batch_size=1
 print(generate_text(keras_model_batch1, 'What of TensorFlow Federated, you ask? '))
 ```

 %% Output

    What of TensorFlow Federated, you ask? Says it with the knitting.
    g of a spy, kept from such hopes to a passenger, whom, one. Piploing And my will be a moment!" said Monsieur Defarge.
    
    
    There were more than twice, went them, and clieding

 %% Cell type:markdown id: tags:

 # Suggested extensions

 This tutorial is just the first step! Here are some ideas for how you might try extending this notebook:
  * Write a more realistic training loop where you sample clients to train on randomly.
  * Use "`.repeat(NUM_EPOCHS)`" on the client datasets to try multiple epochs of local training (e.g., as in [McMahan et. al.](https://arxiv.org/abs/1602.05629)). See also [Federated Learning for Image Classification](federated_learning_for_image_classification.md) which does this.
  * Change the `compile()` command to experiment with using different optimization algorithms on the client.
  * Try the `server_optimizer` argument to `build_federated_averaging_process` to try different algorithms for applying the model updates on the server.
  * Try the `client_weight_fn` argument to to `build_federated_averaging_process` to try different weightings of the clients. The default weights client updates by the number of examples on the client, but you can do e.g. `client_weight_fn=lambda _: tf.constant(1.0)`.