python **kwags错误:函数接受6个位置参数，但给出了8个

Question

当我试图开发梯度下降时，我发现了一个有趣的问题，我无法有效地使用**kwargs。我的功能看起来就像

def gradient_descent(g,x,y,alpha,max_its,w,**kwargs):    
    # switch for verbose
    verbose = True
    if 'verbose' in kwargs:
        verbose = kwargs['verbose']

    # determine num train and batch size
    num_train = y.size()[1]
    batch_size = num_train
    if 'batch_size' in kwargs:
        batch_size = kwargs['batch_size']
    ........

这个错误看起来是：

TypeError                                 Traceback (most recent call last)
<ipython-input-12-f71adb8a241b> in <module>()
  3 w_train = Variable(torch.Tensor(w_init), requires_grad=True)
  4 g = softmax; alpha_choice = 10**(-1); max_its = 100; num_pts = y.size; 
batch_size = 10;
----> 5 weight_hist_2,train_hist_2 = gradient_descent(g,x_train,y_train,alpha_choice,max_its,w_train,num_pts,batch_size,verbose = False)

TypeError: gradient_descent()接受6个位置参数，但给出了8个.

开发这个函数有什么我没注意到的吗？

Answer 1

您的函数签名与您使用它的参数数不匹配：

gradient_descent(g,x,y,alpha,max_its,w,**kwargs)

有6个位置论证g,x,y,alpha,max_its,w，但是，在您的调用中：

gradient_descent(g,x_train,y_train,alpha_choice,max_its,w_train,num_pts,batch_size,verbose = False)

你要给它8 g,x_train,y_train,alpha_choice,max_its,w_train,num_pts,batch_size

我猜您希望使用num_pts作为batch_size参数，因此它将如下所示：

weight_hist_2,train_hist_2 = gradient_descent(
    g,
    x_train,
    y_train,
    alpha_choice,
    max_its,
    w_train,
    batch_size=num_pts,
    verbose = False)