def train_and_evaluate():
# Just for measure the performance of the sentiment analysis model
net = SoundMLP()
net.to(config['device'])
train_loader, test_loader = get_train_loaders()
train_model(net, train_loader, epochs=30, lr=0.0001)
eval_network(net, test_loader)
def train_and_evaluate():
net = get_pretrained_mobile_net(pretrained=True)
net.to(config['device'])
train_loader, test_loader, deploy_loader = get_oasis_dataset_loaders()
for i in range(10):
print(i * 5)
train_model(net, train_loader, epochs=5, lr=0.0001, train_type='image')
eval_network(net, test_loader, train_type='image')
def validate_saved_model():
train_loader = get_deploy_loaders()
model = SoundMLP()
model.to(config['device'])
model.load_neural_model("models/mlp.model")
model.load_neural_space_statistics("models/space_statistics.pickle")
eval_network(model, train_loader)
print("Mean = ", model.mean)
print("Std = ", model.std)
def train_and_evaluate():
net = NeuralTranslator()
net.to(config['device'])
train_loader, test_loader = get_train_loaders(n_clusters=10,
seed=1,
n_sub_classes=10)
train_model(net,
train_loader,
epochs=200,
lr=0.001,
train_type='translator')
eval_network(net, train_loader, train_type='translator')
eval_network(net, test_loader, train_type='translator')
def train_custom_id(imagenet_ids, i):
train_loader = get_deploy_loaders(ids=imagenet_ids[i], n_sub_classes=10)
net = NeuralTranslator()
net.to(config['device'])
train_model(net,
train_loader,
epochs=200,
lr=0.001,
train_type='translator')
eval_network(net, train_loader, train_type='translator')
torch.save(net.state_dict(),
"models/neural_translator_custom_" + str(i) + ".model")
net = NeuralTranslator()
net.to(config['device'])
net.load_state_dict(
torch.load("models/neural_translator_custom_" + str(i) + ".model"))
eval_network(net, train_loader, train_type='translator')
def train_deploy(seed=1):
train_loader = get_deploy_loaders(n_clusters=10,
seed=seed,
n_sub_classes=10)
net = NeuralTranslator()
net.to(config['device'])
train_model(net,
train_loader,
epochs=200,
lr=0.001,
train_type='translator')
eval_network(net, train_loader, train_type='translator')
torch.save(net.state_dict(),
"models/neural_translator_" + str(seed) + ".model")
net = NeuralTranslator()
net.to(config['device'])
net.load_state_dict(
torch.load("models/neural_translator_" + str(seed) + ".model"))
eval_network(net, train_loader, train_type='translator')
def train_deploy():
train_loader, test_loader, deploy_loader = get_oasis_dataset_loaders()
net = get_pretrained_mobile_net(pretrained=True)
net.to(config['device'])
train_model(net, deploy_loader, epochs=50, lr=0.0001, train_type='image')
eval_network(net, test_loader, train_type='image')
train_model(net, deploy_loader, epochs=10, lr=0.00001, train_type='image')
eval_network(net, test_loader, train_type='image')
torch.save(net.state_dict(), "models/image_sentiment.model")
model = get_pretrained_mobile_net(pretrained=True)
model.to(config['device'])
model.load_state_dict(torch.load("models/image_sentiment.model"))
eval_network(model, test_loader, train_type='image')
def train_deploy():
train_loader = get_deploy_loaders()
net = SoundMLP()
net.to(config['device'])
train_model(net, train_loader, epochs=30, lr=0.0001)
eval_network(net, train_loader)
train_model(net, train_loader, epochs=20, lr=0.00001)
eval_network(net, train_loader)
torch.save(net.state_dict(), "models/mlp.model")
model = SoundMLP()
model.to(config['device'])
model.load_state_dict(torch.load("models/mlp.model"))
eval_network(model, train_loader)