如何输入大量数据进行训练 Keras model 以防止 RAM 崩溃？

Question

I'm trying to train a Tensorflow Keras model for a sequential image classification task. The model itself is a simple CNN-RNN model that I've used previously on a classification for 1-D signals, and there's no problem there.

I am having trouble loading the necessary data to train the model on my computer as the RAM gets full and the whole process crashes .

My data looks like this:

(batch, timesteps, height, width, channels) = (batch, 30, 300, 600, 3)

my data pipeline is in this order:

1. glob.glob all the files from one folder into a list
2. load all the data from one file, create an array which is about (50, 30, 300, 600, 3)
3. stack the array from individual file into a continuously growing list using list.append
4. after all the individual file data have been appended, np.vstack to create the final data for training/validation

The above process was okay, but I think appending/vstack is not a good option when doing image processing due to the size of the data.

Is there a way to say have the data saved in a tf.records to reduce overall size? or is there a way to set up a data input pipeline so that data can be loaded in smaller chunks ?

Any help is much appreciated, thank you in advance.

Answer 1

What you need is called DataGenerator

Right now your code probably looks like this:

import numpy as np
from keras.models import Sequential

# Load entire dataset
X, y = np.load('some_training_set_with_labels.npy')

# Design model
model = Sequential()
[...] # Your architecture
model.compile()

# Train model on your dataset
model.fit(x=X, y=y)

Your data-generator will be something like:

import numpy as np
import keras

class DataGenerator(keras.utils.Sequence):
    'Generates data for Keras'
    def __init__(self, list_IDs, labels, batch_size=32, dim=(32,32,32), n_channels=1,
                 n_classes=10, shuffle=True):
        'Initialization'
        self.dim = dim
        self.batch_size = batch_size
        self.labels = labels
        self.list_IDs = list_IDs
        self.n_channels = n_channels
        self.n_classes = n_classes
        self.shuffle = shuffle
        self.on_epoch_end()

    def __len__(self):
        'Denotes the number of batches per epoch'
        return int(np.floor(len(self.list_IDs) / self.batch_size))

    def __getitem__(self, index):
        'Generate one batch of data'
        # Generate indexes of the batch
        indexes = self.indexes[index*self.batch_size:(index+1)*self.batch_size]

        # Find list of IDs
        list_IDs_temp = [self.list_IDs[k] for k in indexes]

        # Generate data
        X, y = self.__data_generation(list_IDs_temp)

        return X, y

    def on_epoch_end(self):
        'Updates indexes after each epoch'
        self.indexes = np.arange(len(self.list_IDs))
        if self.shuffle == True:
            np.random.shuffle(self.indexes)

    def __data_generation(self, list_IDs_temp):
        'Generates data containing batch_size samples' # X : (n_samples, *dim, n_channels)
        # Initialization
        X = np.empty((self.batch_size, *self.dim, self.n_channels))
        y = np.empty((self.batch_size), dtype=int)

        # Generate data
        for i, ID in enumerate(list_IDs_temp):
            # Store sample
            X[i,] = np.load('data/' + ID + '.npy')

            # Store class
            y[i] = self.labels[ID]

        return X, keras.utils.to_categorical(y, num_classes=self.n_classes)

we have to modify our Keras script accordingly so that it accepts the generator that we just created.

import numpy as np

from keras.models import Sequential
from my_classes import DataGenerator

# Parameters
params = {'dim': (32,32,32),
          'batch_size': 64,
          'n_classes': 6,
          'n_channels': 1,
          'shuffle': True}

# Datasets
partition = # IDs
labels = # Labels

# Generators
training_generator = DataGenerator(partition['train'], labels, **params)
validation_generator = DataGenerator(partition['validation'], labels, **params)

# Design model
model = Sequential()
[...] # Architecture
model.compile()

# Train model on dataset
model.fit_generator(generator=training_generator,
                    validation_data=validation_generator,
                    use_multiprocessing=True,
                    workers=4)

Have a look at Stanford University website for more details. It's a bit dated. Have a look at pyimagesearch tutorial for more recent things