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=[
# First generate custom n-shot classification metric
NShotEvaluationCallback(
num_evaluation_tasks,
n_shot_classification,
k_way_classification,
valid,
preprocessor=whiten_downsample,
),
# Then log and checkpoint
CSVLogger(PATH + '/logs/convnet_contrastive_loss.csv'),
ModelCheckpoint(
PATH + '/models/convnet_contrastive_loss.hdf5',
monitor='val_{}-shot_acc'.format(n_shot_classification),
mode='max',
save_best_only=True,
verbose=True)
])
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,
valid,
preprocessor=batch_preprocessor),
SiameseValidator(
batch_gen=valid,
num_tasks=num_evaluation_tasks,
n_shot=1,
k_way=2, # number of speakers sampled
metrics=val_metrics,
monitor='pooled_eer',
mode='min',
preprocessor=batch_preprocessor),
# Then log and checkpoint
CSVLogger(os.path.join(PATH, 'logs/{}.csv'.format(param_str))),
ModelCheckpoint(os.path.join(PATH, 'models/{}.hdf5'.format(param_str)),
monitor='pooled_eer',