如何在 Tensorflow 2 中的嵌入层之后放置 Conv1D 层？

Question

For an evaluation, I need to be able to apply a convolutional layer to text data. So I'm trying to perform sentiment analysis on Amazon reviews. After the Embedding layer, however, the Conv1D layer will not get the required shape.

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
import tensorflow as tf
print(f'Tensorflow version {tf.__version__}')
from tensorflow import keras
from tensorflow.keras.layers import Dense, Conv1D, GlobalAveragePooling1D, Embedding
import tensorflow_datasets as tfds
from tensorflow.keras.models import Model

(train_data, test_data), info = tfds.load('imdb_reviews/subwords8k',
                                          split=[tfds.Split.TRAIN, tfds.Split.TEST],
                                          as_supervised=True, with_info=True)

padded_shapes = ([None], ())

train_dataset = train_data.shuffle(25000).padded_batch(padded_shapes=padded_shapes, batch_size=16)
test_dataset = test_data.shuffle(25000).padded_batch(padded_shapes=padded_shapes, batch_size=16)

n_words = info.features['text'].encoder.vocab_size


class ConvModel(Model):
    def __init__(self):
        super(ConvModel, self).__init__()
        self.embe = Embedding(n_words, output_dim=16)
        self.conv = Conv1D(32, kernel_size=6, activation='elu')
        self.glob = GlobalAveragePooling1D()
        self.dens = Dense(2)

    def call(self, x, training=None, mask=None):
        x = self.embe(x)
        x = self.conv(x)
        x = self.glob(x)
        x = self.dens(x)
        return x

conv = ConvModel()

conv(next(iter(train_data))[0])

ValueError: Input 0 of layer conv1d_25 is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: [163, 16]

How is it possible to achieve this, and if I'm wrong, what is the proper way to use a Conv1D layer to text sequences?

Answer 1

it's conv(next(iter(train_dataset))[0]) and not conv(next(iter(train_data))[0])

the network structure is ok

Answer 2

you have done it so far so good. The last line of the code should be changed. That's all. The parameter should be train_data not the train_dataset.

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
import tensorflow as tf
print(f'Tensorflow version {tf.__version__}')
from tensorflow import keras
from tensorflow.keras.layers import Dense, Conv1D, GlobalAveragePooling1D, Embedding
import tensorflow_datasets as tfds
from tensorflow.keras.models import Model

(train_data, test_data), info = tfds.load('imdb_reviews/subwords8k',
                                          split=[tfds.Split.TRAIN, tfds.Split.TEST],
                                          as_supervised=True, with_info=True)

padded_shapes = ([None], ())

train_dataset = train_data.shuffle(25000).padded_batch(padded_shapes=padded_shapes, batch_size=16)
test_dataset = test_data.shuffle(25000).padded_batch(padded_shapes=padded_shapes, batch_size=16)

n_words = info.features['text'].encoder.vocab_size


class ConvModel(Model):
    def __init__(self):
        super(ConvModel, self).__init__()
        self.embe = Embedding(n_words, output_dim=16)
        self.conv = Conv1D(32, kernel_size=6, activation='elu')
        self.glob = GlobalAveragePooling1D()
        self.dens = Dense(2)

    def call(self, x, training=None, mask=None):
        x = self.embe(x)
        x = self.conv(x)
        x = self.glob(x)
        x = self.dens(x)
        return x

conv = ConvModel()

conv(next(iter(train_data))[0])

Hope you fix the error.

Answer 3

The out_dim of word embedding layer should match with conv1D input filter size. Try changing the out_dim to 32. Proper way: https://machinelearningmastery.com/predict-sentiment-movie-reviews-using-deep-learning/