Data Loader(2024-06-20)

 

 1. 데이터 로더(Data Loader)

* 데이터 양이 많을 때 배치 단위로 학습하는 방법을 제공

 

 

 

 

 2. 손글씨 인식 모델 만들기

 

import torch

import torch.nn as nn

import torch.optim as optim

import matplotlib.pyplot as plt

from sklearn.datasets import load_digits

from sklearn.model_selection import train_test_split

 

device = 'cuda' if torch.cuda.is_available() else 'cpu'

print(device)

 

 

digits = load_digits()

X_data = digits['data']

y_data = digits['target']

print(X_data.shape)

print(y_data.shape)

 

fig, axes = plt.subplots(nrows = 2, ncols = 5, figsize=(14, 8))

for i, ax in enumerate(axes.flatten()) :

  ax.imshow(X_data[i].reshape((8,8)), cmap='gray')

  ax.set_title(y_data[i])

  ax.axis('off')

 

X_data = torch.FloatTensor(X_data)

y_data = torch.LongTensor(y_data)

print(X_data.shape)

print(y_data.shape)

# torch.Size([1797, 64]) torch.Size([1797])

 

X_train, x_test, y_train, y_test = train_test_split(X_data, y_data, test_size = 0.2, random_state = 2024)

print(X_train.shape, y_train.shape)

print(X_test.shape, y_test.shape)

 # torch.Size([1437, 64]) torch.Size([1437]) torch.Size([360, 64]) torch.Size([360])

 

 

loader = torch.utils.data.DataLoader(

    dataset = list(zip(X_train, y_train)),

    batch_size = 64,

    shuffle = True,

    drop_last = False

)

imgs, labels = next(iter(loader))

fig, axex = plt.subplots(nrows= 8, ncols = 8 , figsize=(14, 14))

for ax, img, label in zip(axex.flatten(), imgs, labels) :

  ax.imshow(img.reshape((8,8)), cmap='gray')

  ax.set_title(str(label))

  ax.axis('off')

 

 

 

model = nn.Sequential(

    nn.Linear(64, 10)

)

optimizer = optim.Adam(model.parameters(), lr=0.01)

epochs = 50

for epoch in range(epochs + 1) :

  sum_losses = 0

  sum_accs = 0

  for x_batch, y_batch in loader:

    y_pred = model(x_batch)

    loss = nn.CrossEntropyLoss()(y_pred, y_batch)

    optimizer.zero_grad()

    loss.backward()

    optimizer.step()

    sum_losses += sum_losses + loss

    y_prob = nn.Softmax(1)(y_pred)

    y_pred_index = torch.argmax(y_prob, axis =1)

    acc = (y_batch == y_pred_index).float().sum() / len(y_batch) * 100

    sum_accs = sum_accs + acc

  avg_loss = sum_losses / len(loader)

  avg_acc = sum_accs / len(loader)

  print(f'Epoch {epoch:4d}/{epochs} Loss: {avg_loss:.6f} Accuracy : {avg_acc:.2f}%')