device)) # 計算 loss （注意 prediction 跟 label 必須同時在 CPU 或是 GPU 上）
batch_loss.backward() # 利用 back propagation 算出每個參數的 gradient
optimizer.step() # 以 optimizer 用 gradient 更新參數值
train_acc += np.sum(
np.argmax(train_pred.cpu().data.numpy(), axis=1) ==
data[1].numpy())
train_loss += batch_loss.item()
model.eval()
with torch.no_grad():
for i, data in enumerate(val_loader):
val_pred = model(data[0].to(device))
batch_loss = loss(val_pred, data[1].to(device))
val_acc += np.sum(
np.argmax(val_pred.cpu().data.numpy(), axis=1) ==
data[1].numpy())
val_loss += batch_loss.item()
if val_acc > best_acc:
torch.save(model.state_dict(), 'ckpt.model')
print('saving model with acc {:.3f}'.format(
val_acc / val_set.__len__() * 100))
best_acc = val_acc
#將結果 print 出來
print('[%03d/%03d] %2.2f sec(s) Train Acc: %3.6f Loss: %3.6f | Val Acc: %3.6f loss: %3.6f' % \
(epoch + 1, num_epoch, time.time()-epoch_start_time, \
train_acc/train_set.__len__(), train_loss/train_set.__len__(), val_acc/val_set.__len__(), val_loss/val_set.__len__()))
scheduler.step(val_acc / val_set.__len__())
train_loss += batch_loss.item() # type(batch_loss) == torch.Tensor
model.eval() #same as model.train(False)
with torch.no_grad():
for i, data in enumerate(val_loader):
val_pred = model(data[0].cuda())
batch_loss = loss(val_pred, data[1].cuda())
val_acc += np.sum(
np.argmax(val_pred.cpu().data.numpy(), axis=1) ==
data[1].numpy())
val_loss += batch_loss.item()
print('[%03d/%03d] %2.2f sec(s) Train Acc: %3.6f Loss: %3.6f | Val Acc: %3.6f loss: %3.6f' % \
(epoch + 1, num_epoch, time.time()-epoch_start_time, \
train_acc/train_set.__len__(), train_loss/train_set.__len__(), val_acc/val_set.__len__(), val_loss/val_set.__len__()))
savemodel(model, "./", 1)
'''
#torch.save({'state_dict': model.state_dict()}, 'hw3_parm_b.pkl')
model.eval()
prediction = []
all_pred = []
with torch.no_grad():
for i, data in enumerate(val_loader):
test_pred = model(data[0].cuda())
test_label = np.argmax(test_pred.cpu().data.numpy(), axis=1)
for y in test_label:
prediction.append(y)
for y in test_pred.cpu().data.numpy():
all_pred.append(y)
train_loss += batch_loss.item()
model.eval()
with torch.no_grad():
for i, data in enumerate(val_loader):
val_pred = model(data[0].cuda())
batch_loss = loss(val_pred, data[1].cuda())
val_acc += np.sum(
np.argmax(val_pred.cpu().data.numpy(), axis=1) ==
data[1].numpy())
val_loss += batch_loss.item()
print('[%03d/%03d] %2.2f sec(s) Train Acc: %3.6f Loss: %3.6f | Val Acc: %3.6f loss: %3.6f' % \
(epoch + 1, num_epoch, time.time()-epoch_start_time, \
train_acc/train_set.__len__(), train_loss/train_set.__len__(), val_acc/val_set.__len__(), val_loss/val_set.__len__()))
train_loss_list.append(train_loss / train_set.__len__())
dev_loss_list.append(val_loss / val_set.__len__())
train_acc_list.append(train_acc / train_set.__len__())
dev_acc_list.append(val_acc / val_set.__len__())
if ((val_acc / val_set.__len__()) > val_acc_max) and (
train_acc / train_set.__len__() > train_acc_max):
val_acc_max = val_acc / val_set.__len__()
train_acc_max = train_acc / train_set.__len__()
print("save")
torch.save(model.state_dict(),
"./model_" + str(val_acc / val_set.__len__()))
torch.save(model.state_dict(), "./model_last")
