erro:Expected input batch_size (16) to match target batch_size (32)

1

from tqdm.auto import tqdm

随机种子

torch.manual_seed(42)

循环次数

epoch = 5

for epoch in tqdm(range(epoch)):
print(f’第{epoch}次…')

# 统计训练损失
train_loss = 0

for _,(X,y) in enumerate(train_data_lodael):
   
   # 开始训练
    mode_l.train()
    y_pred = mode_l(X)
     
    # 计算损失
    loss = loss_fn(y_pred,y)
    train_loss += loss
    
    # 优化
    optimizer.zero_grad()
    
    # 后传
    loss.backward()
    
    # 优化stop
    optimizer.step()
    
    if _ % 400 == 0:
        print(f'------------------')

train_loss /= len(train_data_lodael)

# 测试
test_loss,test_acc =0,0
mode_l.eval()
with torch.inference_mode():
    for X_test,y_test in test_data_lodael:
        # 前传
        test_y_pred = mode_l(X_test)
        
        # test
        test_loss += loss_fn(y_pred,y_test)
        
        test_acc += accuracy_fn(y_true=y_test,y_pred=test_y_pred)
        
    # 统计
    test_loss /=len (test_data_lodael)
    
    test_acc /= len(test_data_lodael)
    
print(f'\n训练损失:{train_loss:.4f}| 测试损失:{test_loss:.2f}__测试准确度:{test_acc}')
2

Based on the error message in the title the loss calculation fails due to a shape mismatch.
Could you post your model definition so we could check if you are reshaping the inputs in an invalid way?

PS: you can post code snippets by wrapping them into three backticks ```, which makes debugging easier.