Memory error while training the keras model on 4600000 rows data

Question

I am working on an LSTM-based Encoder-Decoder spelling correction model which is provided with the training data of 4600000 rows. The training file consists of two columns - correct and incorrect sentences. The model was working fine when the data was as small as 200000. But when I increased it the training doesn't go beyond 2 epochs. It sometimes gives the error of terminate called after throwing an instance of std::bad_alloc and sometimes the training stops without any error or warning. I tried using this but it didn't work. Maybe I used it incorrectly.

keras.clear_session() 

I have also tried reducing the value of latent_dim and batch_size to 128,64,32,16,8,4,1 but none of them worked for such large data. Also since the data was huge so I replaced the value of

steps_per_epoch = train_samples//batch_size

to

steps_per_epoch = 2000

I cleared the cache to free up ram but still, the training doesn't complete. Can someone suggest a way to train my model?

def generate_batch(X = X_train, y = y_train, batch_size = 128):
    # Generate a batch of data 
    while True:
        for j in range(0, len(X), batch_size):
            encoder_input_data = np.zeros((batch_size, max_length_src),dtype='float32')
            decoder_input_data = np.zeros((batch_size, max_length_tar),dtype='float32')
            decoder_target_data = np.zeros((batch_size, max_length_tar, num_decoder_tokens),dtype='float32')
            for i, (input_text, target_text) in enumerate(zip(X[j:j+batch_size], y[j:j+batch_size])):
                for t, word in enumerate(input_text.split()):
                    encoder_input_data[i, t] = input_token_index[word] # encoder input seq
                for t, word in enumerate(target_text.split()):
                    if t<len(target_text.split())-1:
                        decoder_input_data[i, t] = target_token_index[word] # decoder input seq
                    if t>0:
                        # decoder target sequence (one hot encoded)
                        # does not include the START_ token
                        # Offset by one timestep
                        decoder_target_data[i, t - 1, target_token_index[word]] = 1.
            yield([encoder_input_data, decoder_input_data], decoder_target_data)

latent_dim = 50

# Encoder
encoder_inputs = Input(shape=(None,))
enc_emb =  Embedding(num_encoder_tokens+1, latent_dim, mask_zero = True)(encoder_inputs)
encoder_lstm = LSTM(latent_dim, return_state=True)
encoder_outputs, state_h, state_c = encoder_lstm(enc_emb)
# We discard `encoder_outputs` and only keep the states.
encoder_states = [state_h, state_c]

# Set up the decoder, using `encoder_states` as initial state.
decoder_inputs = Input(shape=(None,))
dec_emb_layer = Embedding(num_decoder_tokens, latent_dim, mask_zero = True)
dec_emb = dec_emb_layer(decoder_inputs)
# We set up our decoder to return full output sequences,
# and to return internal states as well. We don't use the
# return states in the training model, but we will use them in inference.
decoder_lstm = LSTM(latent_dim, return_sequences=True, return_state=True)
decoder_outputs, _, _ = decoder_lstm(dec_emb,
                                     initial_state=encoder_states)
decoder_dense = Dense(num_decoder_tokens, activation='softmax')
decoder_outputs = decoder_dense(decoder_outputs)

# Define the model that will turn
# `encoder_input_data` & `decoder_input_data` into `decoder_target_data`
model = Model([encoder_inputs, decoder_inputs], decoder_outputs)

model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])

train_samples = len(X_train)
val_samples = len(X_test)
batch_size = 128
epochs = 50

from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint
from keras.layers import Conv2D, MaxPooling2D
from keras import backend as K
from keras.callbacks import ModelCheckpoint

keras_callbacks   = [
      EarlyStopping(monitor ="val_loss", mode ="min", patience = 5, restore_best_weights = True),
      ModelCheckpoint('checkpoints.hdf5', monitor='val_loss', verbose=1, save_best_only=True, mode='min', save_freq=1)
]

model.fit_generator(generator = generate_batch(X_train, y_train, batch_size = batch_size),
                    #steps_per_epoch = train_samples//batch_size,
                    steps_per_epoch = 2000,
                    epochs=epochs,
                    verbose=1,
                    validation_data = generate_batch(X_test, y_test, batch_size = batch_size),
                    validation_steps = val_samples//batch_size,
                    callbacks=keras_callbacks)

model.save_weights('weights.h5')

Answer 1

The memory error occurs because you're using "big" global scope variable input_token_index in generate_batch function. This variable will be copied many times in memory to generate your data.

However, instead of fixing this particular problem, I would suggest you using native TF functionality for text tokenisation, vectorisation and batching instead of writing your own implementations.

You can find more info about text tokenisation and vectorisation here - official Tensorflow tutorial ). Specifically, you can utilise Tensorflow text vectorisation functionality, which combines tokenisation and padding. Alternatively, you can use more mature Tokeniser and general pad_sequences

Creating batches is a super basic and general task for SGD type of training so don't reinvent the wheel and just use model.fit , it will automatically batch your data for free.