def run_sop(hyper, results_path):
tf.random.set_seed(seed=hyper['seed'])
data = load_mnist_sop_data(batch_n=hyper['batch_size'])
train_dataset, test_dataset = data
sop_optimizer = setup_sop_optimizer(hyper=hyper)
model_type = sop_optimizer.model.model_type
log_path = results_path + f'/loss_{model_type}.log'
logger = setup_logger(log_file_name=append_timestamp_to_file(
file_name=log_path, termination='.log'),
logger_name=model_type + str(hyper['seed']))
log_all_hyperparameters(hyper=hyper, logger=logger)
save_hyper(hyper)
train_sop(sop_optimizer=sop_optimizer,
hyper=hyper,
train_dataset=train_dataset,
test_dataset=test_dataset,
logger=logger)
def test_in_mnist_sample():
batch_n, epochs, width, height, rgb = 4, 3, 14, 28, 1
hyper = {
'width_height': (width, height, rgb),
'model_type': '',
'units_per_layer': 240,
'batch_size': batch_n,
'learning_rate': 0.0003,
'weight_decay': 1.e-3,
'epochs': epochs,
'iter_per_epoch': 10,
'test_sample_size': 1,
'temp': tf.constant(0.1)
}
results_path = './Log/'
data = load_mnist_sop_data(batch_n=batch_n, run_with_sample=True)
models = ['GS', 'IGR_I', 'IGR_Planar']
# models = ['GS', 'IGR_I']
for model in models:
hyper['model_type'] = model
run_sop(hyper=hyper, results_path=results_path, data=data)
def test_in_mnist_sample(self):
batch_n = 6
epochs = 5
width = 14
height = 28
hyper = {
'width_height': (width, height, 1),
'model_type': 'GS',
'batch_size': batch_n,
'learning_rate': 0.001,
'epochs': epochs,
'iter_per_epoch': 10,
'temp': tf.constant(0.1)
}
results_path = './Log/'
data = load_mnist_sop_data(batch_n=batch_n,
epochs=epochs,
run_with_sample=True)
train_dataset, test_dataset = data
run_sop(hyper=hyper, results_path=results_path, data=data)
hyper['model_type'] = 'IGR'
run_sop(hyper=hyper, results_path=results_path, data=data)
select_case = 2
run_with_sample = False
# samples_n = 1 * int(1.e3)
samples_n = 1 * int(1.e3)
hyper_file, weights_file = 'hyper.pkl', 'w.h5'
model_type = models[select_case]['model_type']
path_to_trained_models += models[select_case]['model_dir'] + '/'
with open(file=path_to_trained_models + hyper_file, mode='rb') as f:
hyper = pickle.load(f)
batch_n = hyper['batch_size']
hyper['test_sample_size'] = samples_n
tf.random.set_seed(seed=hyper['seed'])
data = load_mnist_sop_data(batch_n=hyper['batch_size'],
run_with_sample=run_with_sample)
train_dataset, test_dataset = data
epoch = hyper['epochs']
sop_optimizer = setup_sop_optimizer(hyper=hyper)
for x in train_dataset:
x_upper = x[:, :14, :, :]
break
sop_optimizer.batch_n = batch_n
aux = sop_optimizer.model.call(x_upper)
sop_optimizer.model.load_weights(filepath=path_to_trained_models +
weights_file)
test_loss_mean = tf.keras.metrics.Mean()
for x in test_dataset:
loss = evaluate_loss_on_batch(x, sop_optimizer, hyper)
test_loss_mean(loss)
from Utils.load_data import load_mnist_sop_data
from Models.SOP import SOP
from Models.SOPOptimizer import viz_reconstruction
import tensorflow as tf
model_type = 'GS'
hyper = {
'width_height': (14, 28, 1),
'model_type': model_type,
'batch_size': 64,
'learning_rate': 0.0003,
'epochs': 100,
'iter_per_epoch': 937,
'temp': tf.constant(0.67)
}
data = load_mnist_sop_data(batch_n=hyper['batch_size'], epochs=hyper['epochs'])
train, test = data
model = SOP(hyper=hyper)
results_file = './Log/model_weights_GS.h5'
shape = (hyper['batch_size'], ) + hyper['width_height']
shape = (64, 14, 28, 1)
model.build(input_shape=shape)
model.load_weights(filepath=results_file)
for x_test in test.take(10):
images = x_test
viz_reconstruction(test_image=images, model=model)
from Utils.load_data import load_mnist_sop_data
from Models.SOP import SOP
from Models.SOPOptimizer import viz_reconstruction
import tensorflow as tf
model_type = 'GS'
hyper = {
'width_height': (14, 28, 1),
'model_type': model_type,
'batch_size': 64,
'learning_rate': 0.0003,
'epochs': 100,
'iter_per_epoch': 937,
'temp': tf.constant(0.67)
}
data = load_mnist_sop_data(batch_n=hyper['batch_size'])
train, test = data
model = SOP(hyper=hyper)
results_file = './Log/model_weights_GS.h5'
shape = (hyper['batch_size'], ) + hyper['width_height']
model.build(input_shape=shape)
model.load_weights(filepath=results_file)
images = 0
for x_test in test.take(10):
images = x_test
viz_reconstruction(test_image=images, model=model)