nn.Softmax

import torch
from torch import nn

a = torch.tensor( [1,2,4,3,5,5], dtype=torch.float32)
a = a.view(3,2)
print(a)
print(a.shape)

# tensor([[1., 2.],
#         [4., 3.],
#         [5., 5.]])
# torch.Size([3, 2])

softmax = nn.Softmax(dim=0)
softmax_a = softmax(a)
print(softmax_a)

# tensor([[0.0132, 0.0420],
#         [0.2654, 0.1142],
#         [0.7214, 0.8438]])

: dim=0 ) row
: row(행)들끼리 Softmax 계산

softmax_a.sum(dim=0) 
tensor([1., 1.])

: row끼리 softmax연산 했으므로 row들의 합 = 1

model parameter optimize

• loss function으로 loss 계산
• 역전파
  - optimizer.zero_grad()
  : backward()함수는 새로운 gradient를 기존 gradient에 (+=)누적시키므로 0으로 초기화해줘야 한다.
  - loss.backward()
  : loss function의 미분값, gradients 저장
  - optimizer.step()
  : gradients를 바탕으로 parameter 조정

  • autograd
    loss.backward()
    h : tensor 변수(w)가 포함된 수식
    h.backward() : h를 w로 미분한 값
    h.backward()의 결과값 -> w.grad에 저장됨.

def train_loop(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)
    for batch, (X, y) in enumerate(dataloader):
        # 예측(prediction)과 손실(loss) 계산
        pred = model(X)
        loss = loss_fn(pred, y)

        # 역전파
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        if batch % 100 == 0:
            loss, current = loss.item(), batch * len(X)
            print(f"loss: {loss:>7f}  [{current:>5d}/{size:>5d}]")
            
def test_loop(dataloader, model, loss_fn):
    size = len(dataloader.dataset)
    num_batches = len(dataloader)
    test_loss, correct = 0, 0

    with torch.no_grad():
        for X, y in dataloader:
            pred = model(X)
            test_loss += loss_fn(pred, y).item()
            correct += (pred.argmax(1) == y).type(torch.float).sum().item()

    test_loss /= num_batches
    correct /= size
    print(f"Test Error: \n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f} \n")   

loss_fn = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)

epochs = 10
for t in range(epochs):
    print(f"Epoch {t+1}\n-------------------------------")
    train_loop(train_dataloader, model, loss_fn, optimizer)
    test_loop(test_dataloader, model, loss_fn)
print("Done!")

참고

: https://velog.io/@olxtar/nn.Softmax