mirror of
https://github.com/kuhyx/WUT_Computer_Science.git
synced 2026-11-20 12:12:23 +01:00
feat: add constants
This commit is contained in:
parent
d8088a38b4
commit
9f271d07a6
1
lab5/.gitignore
vendored
Normal file
1
lab5/.gitignore
vendored
Normal file
@ -0,0 +1 @@
|
|||||||
|
data/
|
||||||
@ -5,15 +5,19 @@ from torch import optim
|
|||||||
from torchvision import datasets, transforms
|
from torchvision import datasets, transforms
|
||||||
import matplotlib.pyplot as plt
|
import matplotlib.pyplot as plt
|
||||||
|
|
||||||
|
# Global Constants
|
||||||
|
LEARNING_RATE = 0.001
|
||||||
|
BATCH_SIZE = 64
|
||||||
|
NUM_HIDDEN_LAYERS = 2
|
||||||
|
WIDTH = 128
|
||||||
|
OPTIMIZER_TYPE = 'Adam'
|
||||||
|
|
||||||
|
|
||||||
def set_hyperparameters():
|
def set_hyperparameters():
|
||||||
""" sets hyperparameters used throughout the network """
|
""" sets hyperparameters used throughout the network """
|
||||||
return {
|
return {
|
||||||
"learning_rate": 0.001,
|
"num_epochs": 5,
|
||||||
"batch_size": 64,
|
"init_input_size": 28 * 28, # MNIST images are 28x28 pixels
|
||||||
"num_epochs": 2,
|
|
||||||
"input_size": 28 * 28, # MNIST images are 28x28 pixels
|
|
||||||
"hidden_size": 128,
|
|
||||||
"num_classes": 10,
|
"num_classes": 10,
|
||||||
}
|
}
|
||||||
|
|
||||||
@ -29,13 +33,13 @@ def load_datasets():
|
|||||||
return train_dataset, test_dataset
|
return train_dataset, test_dataset
|
||||||
|
|
||||||
|
|
||||||
def create_data_loaders(train_dataset, test_dataset, hyperparameters):
|
def create_data_loaders(train_dataset, test_dataset):
|
||||||
""" Create train and test data loaders """
|
""" Create train and test data loaders """
|
||||||
train_loader = torch.utils.data.DataLoader(
|
train_loader = torch.utils.data.DataLoader(
|
||||||
dataset=train_dataset, batch_size=hyperparameters["batch_size"], shuffle=True
|
dataset=train_dataset, batch_size=BATCH_SIZE, shuffle=True
|
||||||
)
|
)
|
||||||
test_loader = torch.utils.data.DataLoader(
|
test_loader = torch.utils.data.DataLoader(
|
||||||
dataset=test_dataset, batch_size=hyperparameters["batch_size"], shuffle=False
|
dataset=test_dataset, batch_size=BATCH_SIZE, shuffle=False
|
||||||
)
|
)
|
||||||
return train_loader, test_loader
|
return train_loader, test_loader
|
||||||
|
|
||||||
@ -48,16 +52,31 @@ val_acc_values = []
|
|||||||
|
|
||||||
def define_model(hyperparameters):
|
def define_model(hyperparameters):
|
||||||
""" Define the multilayer perceptron training_parameters['model'] """
|
""" Define the multilayer perceptron training_parameters['model'] """
|
||||||
model = nn.Sequential(
|
# Define the multilayer perceptron model
|
||||||
nn.Linear(hyperparameters["input_size"],
|
model = nn.Sequential()
|
||||||
hyperparameters["hidden_size"]),
|
model.add_module('flatten', nn.Flatten())
|
||||||
nn.ReLU(),
|
input_size = hyperparameters['init_input_size']
|
||||||
nn.Linear(hyperparameters["hidden_size"],
|
for i in range(NUM_HIDDEN_LAYERS):
|
||||||
hyperparameters["num_classes"]),
|
model.add_module(f'linear{i}', nn.Linear(input_size, WIDTH))
|
||||||
)
|
model.add_module(f'relu{i}', nn.ReLU())
|
||||||
|
input_size = WIDTH
|
||||||
|
model.add_module('output', nn.Linear(
|
||||||
|
input_size, hyperparameters['num_classes']))
|
||||||
return model
|
return model
|
||||||
|
|
||||||
|
|
||||||
|
def get_optimizer(model):
|
||||||
|
""" Return optimizer function """
|
||||||
|
if OPTIMIZER_TYPE == 'SGD':
|
||||||
|
return optim.SGD(model.parameters(), lr=LEARNING_RATE)
|
||||||
|
elif OPTIMIZER_TYPE == 'SGD_Momentum':
|
||||||
|
return optim.SGD(model.parameters(), lr=LEARNING_RATE, momentum=0.9)
|
||||||
|
elif OPTIMIZER_TYPE == 'Adam':
|
||||||
|
return optim.Adam(model.parameters(), lr=LEARNING_RATE)
|
||||||
|
else:
|
||||||
|
raise ValueError("Unsupported optimizer type!")
|
||||||
|
|
||||||
|
|
||||||
def initial_configuration():
|
def initial_configuration():
|
||||||
"""
|
"""
|
||||||
Perform all operations needed for training network
|
Perform all operations needed for training network
|
||||||
@ -68,14 +87,12 @@ def initial_configuration():
|
|||||||
# Load MNIST dataset and apply transformations
|
# Load MNIST dataset and apply transformations
|
||||||
train_dataset, test_dataset = load_datasets()
|
train_dataset, test_dataset = load_datasets()
|
||||||
train_loader, test_loader = create_data_loaders(
|
train_loader, test_loader = create_data_loaders(
|
||||||
train_dataset, test_dataset, hyperparameters
|
train_dataset, test_dataset)
|
||||||
)
|
|
||||||
model = define_model(hyperparameters)
|
model = define_model(hyperparameters)
|
||||||
# Loss function
|
# Loss function
|
||||||
criterion = nn.CrossEntropyLoss()
|
criterion = nn.CrossEntropyLoss()
|
||||||
# training_parameters['optimizer']
|
# training_parameters['optimizer']
|
||||||
optimizer = optim.Adam(
|
optimizer = get_optimizer(model)
|
||||||
model.parameters(), lr=hyperparameters["learning_rate"])
|
|
||||||
return hyperparameters, train_loader, test_loader, model, criterion, optimizer
|
return hyperparameters, train_loader, test_loader, model, criterion, optimizer
|
||||||
|
|
||||||
|
|
||||||
@ -99,7 +116,7 @@ def single_train_iteration(
|
|||||||
print(
|
print(
|
||||||
f'''
|
f'''
|
||||||
Epoch [{epoch+1}/{training_parameters['hyperparameters']["num_epochs"]}],
|
Epoch [{epoch+1}/{training_parameters['hyperparameters']["num_epochs"]}],
|
||||||
Step [{batch_idx+1}/ {len(training_parameters['loaders']['train_loader'])}],
|
Step [{batch_idx+1}/{len(training_parameters['loaders']['train_loader'])}],
|
||||||
Loss: {loss.item():.4f}
|
Loss: {loss.item():.4f}
|
||||||
'''
|
'''
|
||||||
)
|
)
|
||||||
|
|||||||
Loading…
Reference in New Issue
Block a user