KerryHalupka

KerryHalupka

KerryHalupka

KerryHalupka

KerryHalupka

KerryHalupka

model.load_weights(ckpt_dir)
for inputs in test_gen:
    outputs = model(inputs, training=False)
    break

KerryHalupka

#load model weights
model.load_state_dict(torch.load(ckpt_dir))
    
for inputs, targets in test_gen:
    # use GPU if available
    inputs = inputs.to(device)
    inputs = inputs.float()
    outputs = model(inputs)
    break

KerryHalupka

learning_rate = 0.0001

#get mobilenet
model = torchvision.models.mobilenet_v2(pretrained=True)

# freeze the feature extraction convolutional layers
for param in model.parameters():
    param.requires_grad = False

# define a classification layer

KerryHalupka

from torch.utils.tensorboard import SummaryWriter

# Setup tensorboard
file_writer = SummaryWriter(log_dir=log_dir + '/metrics')
best_val_acc = 0 # for model check pointing
# Epoch loop
for epoch in range(1, num_epoch + 1):
    start_time = timer()
    # Reset metrics
    train_loss = 0.0

KerryHalupka

learning_rate = 0.0001
# get just the feature extraction layers of mobilenet
base_model = tf.keras.applications.MobileNetV2(
    include_top=False, weights='imagenet', input_shape=(im_size, im_size, 3))

# freeze the feature extractor convolutional layers
base_model.trainable = False

# define a classification layer on top
global_average_layer = tf.keras.layers.GlobalAveragePooling2D()