 View on TensorFlow.org
|
 View source on GitHub
|
 Download notebook
|
This text classification tutorial trains a recurrent neural network on the IMDB large movie review dataset for sentiment analysis.
Setup
import tensorflow_datasets as tfds
import tensorflow as tf
Import matplotlib and create a helper function to plot graphs:
import matplotlib.pyplot as plt
def plot_graphs(history, metric):
plt.plot(history.history[metric])
plt.plot(history.history['val_'+metric], '')
plt.xlabel("Epochs")
plt.ylabel(metric)
plt.legend([metric, 'val_'+metric])
plt.show()
Setup input pipeline
The IMDB large movie review dataset is a binary classification dataset—all the reviews have either a positive or negative sentiment.
Download the dataset using TFDS.
dataset, info = tfds.load('imdb_reviews/subwords8k', with_info=True,
as_supervised=True)
train_dataset, test_dataset = dataset['train'], dataset['test']
WARNING:absl:TFDS datasets with text encoding are deprecated and will be removed in a future version. Instead, you should use the plain text version and tokenize the text using `tensorflow_text` (See: https://www.tensorflow.org/tutorials/tensorflow_text/intro#tfdata_example)
The dataset info includes the encoder (a tfds.features.text.SubwordTextEncoder).
encoder = info.features['text'].encoder
print('Vocabulary size: {}'.format(encoder.vocab_size))
Vocabulary size: 8185
This text encoder will reversibly encode any string, falling back to byte-encoding if necessary.
sample_string = 'Hello TensorFlow.'
encoded_string = encoder.encode(sample_string)
print('Encoded string is {}'.format(encoded_string))
original_string = encoder.decode(encoded_string)
print('The original string: "{}"'.format(original_string))
Encoded string is [4025, 222, 6307, 2327, 4043, 2120, 7975] The original string: "Hello TensorFlow."
assert original_string == sample_string
for index in encoded_string:
print('{} ----> {}'.format(index, encoder.decode([index])))
4025 ----> Hell 222 ----> o 6307 ----> Ten 2327 ----> sor 4043 ----> Fl 2120 ----> ow 7975 ----> .
Prepare the data for training
Next create batches of these encoded strings. Use the padded_batch method to zero-pad the sequences to the length of the longest string in the batch:
BUFFER_SIZE = 10000
BATCH_SIZE = 64
train_dataset = train_dataset.shuffle(BUFFER_SIZE)
train_dataset = train_dataset.padded_batch(BATCH_SIZE)
test_dataset = test_dataset.padded_batch(BATCH_SIZE)
Create the model
Build a tf.keras.Sequential model and start with an embedding layer. An embedding layer stores one vector per word. When called, it converts the sequences of word indices to sequences of vectors. These vectors are trainable. After training (on enough data), words with similar meanings often have similar vectors.
This index-lookup is much more efficient than the equivalent operation of passing a one-hot encoded vector through a tf.keras.layers.Dense layer.
A recurrent neural network (RNN) processes sequence input by iterating through the elements. RNNs pass the outputs from one timestep to their input—and then to the next.
The tf.keras.layers.Bidirectional wrapper can also be used with an RNN layer. This propagates the input forward and backwards through the RNN layer and then concatenates the output. This helps the RNN to learn long range dependencies.
model = tf.keras.Sequential([
tf.keras.layers.Embedding(encoder.vocab_size, 64),
tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(64)),
tf.keras.layers.Dense(64, activation='relu'),
tf.keras.layers.Dense(1)
])
Please note that we choose to Keras sequential model here since all the layers in the model only have single input and produce single output. In case you want to use stateful RNN layer, you might want to build your model with Keras functional API or model subclassing so that you can retrieve and reuse the RNN layer states. Please check Keras RNN guide for more details.
Compile the Keras model to configure the training process:
model.compile(loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
optimizer=tf.keras.optimizers.Adam(1e-4),
metrics=['accuracy'])
Train the model
history = model.fit(train_dataset, epochs=10,
validation_data=test_dataset,
validation_steps=30)
Epoch 1/10 391/391 [==============================] - 42s 107ms/step - loss: 0.6714 - accuracy: 0.5397 - val_loss: 0.5508 - val_accuracy: 0.7661 Epoch 2/10 391/391 [==============================] - 42s 106ms/step - loss: 0.3864 - accuracy: 0.8256 - val_loss: 0.3704 - val_accuracy: 0.8203 Epoch 3/10 391/391 [==============================] - 42s 108ms/step - loss: 0.2683 - accuracy: 0.8948 - val_loss: 0.3526 - val_accuracy: 0.8375 Epoch 4/10 391/391 [==============================] - 42s 107ms/step - loss: 0.2247 - accuracy: 0.9149 - val_loss: 0.3221 - val_accuracy: 0.8641 Epoch 5/10 391/391 [==============================] - 42s 107ms/step - loss: 0.1886 - accuracy: 0.9323 - val_loss: 0.3283 - val_accuracy: 0.8562 Epoch 6/10 391/391 [==============================] - 42s 106ms/step - loss: 0.1670 - accuracy: 0.9409 - val_loss: 0.3456 - val_accuracy: 0.8672 Epoch 7/10 391/391 [==============================] - 42s 108ms/step - loss: 0.1515 - accuracy: 0.9468 - val_loss: 0.4069 - val_accuracy: 0.8646 Epoch 8/10 391/391 [==============================] - 42s 107ms/step - loss: 0.1396 - accuracy: 0.9507 - val_loss: 0.3695 - val_accuracy: 0.8542 Epoch 9/10 391/391 [==============================] - 42s 108ms/step - loss: 0.1244 - accuracy: 0.9591 - val_loss: 0.4349 - val_accuracy: 0.8557 Epoch 10/10 391/391 [==============================] - 42s 108ms/step - loss: 0.1150 - accuracy: 0.9615 - val_loss: 0.4144 - val_accuracy: 0.8552
test_loss, test_acc = model.evaluate(test_dataset)
print('Test Loss: {}'.format(test_loss))
print('Test Accuracy: {}'.format(test_acc))
391/391 [==============================] - 18s 45ms/step - loss: 0.4298 - accuracy: 0.8500 Test Loss: 0.4298378825187683 Test Accuracy: 0.8500400185585022
The above model does not mask the padding applied to the sequences. This can lead to skew if trained on padded sequences and test on un-padded sequences. Ideally you would use masking to avoid this, but as you can see below it only have a small effect on the output.
If the prediction is >= 0.5, it is positive else it is negative.
def pad_to_size(vec, size):
zeros = [0] * (size - len(vec))
vec.extend(zeros)
return vec
def sample_predict(sample_pred_text, pad):
encoded_sample_pred_text = encoder.encode(sample_pred_text)
if pad:
encoded_sample_pred_text = pad_to_size(encoded_sample_pred_text, 64)
encoded_sample_pred_text = tf.cast(encoded_sample_pred_text, tf.float32)
predictions = model.predict(tf.expand_dims(encoded_sample_pred_text, 0))
return (predictions)
# predict on a sample text without padding.
sample_pred_text = ('The movie was cool. The animation and the graphics '
'were out of this world. I would recommend this movie.')
predictions = sample_predict(sample_pred_text, pad=False)
print(predictions)
[[-0.29123712]]
# predict on a sample text with padding
sample_pred_text = ('The movie was cool. The animation and the graphics '
'were out of this world. I would recommend this movie.')
predictions = sample_predict(sample_pred_text, pad=True)
print(predictions)
[[-0.36919916]]
plot_graphs(history, 'accuracy')
plot_graphs(history, 'loss')
Stack two or more LSTM layers
Keras recurrent layers have two available modes that are controlled by the return_sequences constructor argument:
- Return either the full sequences of successive outputs for each timestep (a 3D tensor of shape
(batch_size, timesteps, output_features)). - Return only the last output for each input sequence (a 2D tensor of shape (batch_size, output_features)).
model = tf.keras.Sequential([
tf.keras.layers.Embedding(encoder.vocab_size, 64),
tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(64, return_sequences=True)),
tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(32)),
tf.keras.layers.Dense(64, activation='relu'),
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(1)
])
model.compile(loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
optimizer=tf.keras.optimizers.Adam(1e-4),
metrics=['accuracy'])
history = model.fit(train_dataset, epochs=10,
validation_data=test_dataset,
validation_steps=30)
Epoch 1/10 391/391 [==============================] - 73s 188ms/step - loss: 0.6218 - accuracy: 0.5949 - val_loss: 0.4266 - val_accuracy: 0.8156 Epoch 2/10 391/391 [==============================] - 72s 185ms/step - loss: 0.3395 - accuracy: 0.8630 - val_loss: 0.3570 - val_accuracy: 0.8490 Epoch 3/10 391/391 [==============================] - 73s 186ms/step - loss: 0.2509 - accuracy: 0.9078 - val_loss: 0.3568 - val_accuracy: 0.8484 Epoch 4/10 391/391 [==============================] - 72s 185ms/step - loss: 0.2064 - accuracy: 0.9293 - val_loss: 0.3575 - val_accuracy: 0.8573 Epoch 5/10 391/391 [==============================] - 72s 185ms/step - loss: 0.1778 - accuracy: 0.9429 - val_loss: 0.3942 - val_accuracy: 0.8625 Epoch 6/10 391/391 [==============================] - 75s 193ms/step - loss: 0.1615 - accuracy: 0.9493 - val_loss: 0.4104 - val_accuracy: 0.8531 Epoch 7/10 391/391 [==============================] - 75s 191ms/step - loss: 0.1424 - accuracy: 0.9571 - val_loss: 0.4190 - val_accuracy: 0.8536 Epoch 8/10 391/391 [==============================] - 75s 192ms/step - loss: 0.1252 - accuracy: 0.9643 - val_loss: 0.4940 - val_accuracy: 0.8469 Epoch 9/10 391/391 [==============================] - 75s 192ms/step - loss: 0.1122 - accuracy: 0.9686 - val_loss: 0.5188 - val_accuracy: 0.8484 Epoch 10/10 391/391 [==============================] - 73s 187ms/step - loss: 0.0961 - accuracy: 0.9746 - val_loss: 0.5247 - val_accuracy: 0.8547
test_loss, test_acc = model.evaluate(test_dataset)
print('Test Loss: {}'.format(test_loss))
print('Test Accuracy: {}'.format(test_acc))
391/391 [==============================] - 32s 82ms/step - loss: 0.5032 - accuracy: 0.8538 Test Loss: 0.5032243132591248 Test Accuracy: 0.8537600040435791
# predict on a sample text without padding.
sample_pred_text = ('The movie was not good. The animation and the graphics '
'were terrible. I would not recommend this movie.')
predictions = sample_predict(sample_pred_text, pad=False)
print(predictions)
[[-2.6426256]]
# predict on a sample text with padding
sample_pred_text = ('The movie was not good. The animation and the graphics '
'were terrible. I would not recommend this movie.')
predictions = sample_predict(sample_pred_text, pad=True)
print(predictions)
[[-3.7189987]]
plot_graphs(history, 'accuracy')
plot_graphs(history, 'loss')
Check out other existing recurrent layers such as GRU layers.
If you're interestied in building custom RNNs, see the Keras RNN Guide.