###################
# Create datasets #
###################
train = LibriSpeechDataset(training_set, n_seconds)
valid = LibriSpeechDataset(validation_set, n_seconds, stochastic=False)
train_generator = (whiten_downsample(batch)
for batch in train.yield_verification_batches(batchsize))
valid_generator = (whiten_downsample(batch)
for batch in valid.yield_verification_batches(batchsize))
################
# Define model #
################
encoder = get_baseline_convolutional_encoder(model_n_filters,
model_embedding_dimension)
siamese = build_siamese_net(encoder, (input_length, 1))
opt = Adam(clipnorm=1.)
siamese.compile(loss=contrastive_loss, optimizer=opt, metrics=['accuracy'])
#################
# Training Loop #
#################
siamese.fit_generator(
generator=train_generator,
steps_per_epoch=evaluate_every_n_batches,
validation_data=valid_generator,
validation_steps=100,
epochs=num_epochs,
workers=multiprocessing.cpu_count(),
use_multiprocessing=True,
callbacks=[
pad=pad)
batch_preprocessor = BatchPreProcessor('siamese',
preprocess_instances(downsampling))
train_generator = (batch_preprocessor(batch)
for batch in train.yield_verification_batches(batchsize))
valid_generator = (batch_preprocessor(batch)
for batch in valid.yield_verification_batches(batchsize))
################
# Define model #
################
encoder = get_baseline_convolutional_encoder(filters,
embedding_dimension,
dropout=dropout)
siamese = build_siamese_net(encoder, (input_length, 1),
distance_metric='uniform_euclidean')
opt = Adam(clipnorm=1.)
siamese.compile(loss='binary_crossentropy',
optimizer=opt,
metrics=['accuracy'])
# plot_model(siamese, show_shapes=True, to_file=PATH + '/plots/siamese.png')
print(siamese.summary())
#################
# Training Loop #
#################
callbacks = [
# First generate custom n-shot classification metric
NShotEvaluationCallback(num_evaluation_tasks,
n_shot_classification,
k_way_classification,
