Autograd.grad() 用于 pytorch 中的 Tensor [英] Autograd.grad() for Tensor in pytorch

问题描述

我想计算网络中两个张量之间的梯度.输入 X 张量(批量大小 x m)通过一组卷积层发送给我，然后输出 Y 张量(批量大小 x n).

I want to compute the gradient between two tensors in a net. The input X tensor (batch size x m) is sent through a set of convolutional layers which give me back and output Y tensor(batch size x n).

我正在创建一个新的损失，我想知道 Y w.r.t. 的梯度.X. tensorflow 中的一些东西:

I’m creating a new loss and I would like to know the gradient of Y w.r.t. X. Something that in tensorflow would be like:

tf.gradients(ys=Y, xs=X)

不幸的是，我一直在使用 torch.autograd.grad() 进行测试，但我不知道如何去做.我收到如下错误:RunTimeerror: grad can be hidden created only for scalar output".

Unfortunately, I’ve been making tests with torch.autograd.grad(), but I could not figure out how to do it. I get errors like: "RunTimeerror: grad can be implicitly created only for scalar outputs".

如果我想知道 Y w.r.t. 的梯度，torch.autograd.grad() 中的输入应该是什么?X?

What should be the inputs in torch.autograd.grad() if I want to know the gradient of Y w.r.t. X?

推荐答案

让我们从简单的工作示例开始，使用简单的损失函数和正则向后.我们将构建简短的计算图并对其进行一些梯度计算.

Let's start from simple working example with plain loss function and regular backward. We will build short computational graph and do some grad computations on it.

代码:

import torch
from torch.autograd import grad
import torch.nn as nn


# Create some dummy data.
x = torch.ones(2, 2, requires_grad=True)
gt = torch.ones_like(x) * 16 - 0.5  # "ground-truths" 

# We will use MSELoss as an example.
loss_fn = nn.MSELoss()

# Do some computations.
v = x + 2
y = v ** 2

# Compute loss.
loss = loss_fn(y, gt)

print(f'Loss: {loss}')

# Now compute gradients:
d_loss_dx = grad(outputs=loss, inputs=x)
print(f'dloss/dx:
 {d_loss_dx}')

输出:

Loss: 42.25
dloss/dx:
(tensor([[-19.5000, -19.5000], [-19.5000, -19.5000]]),)

好的，这有效！现在让我们尝试重现错误只能为标量输出隐式创建 grad".如您所见，前面示例中的 loss 是一个标量.backward() 和 grad() 默认处理单个标量值:loss.backward(torch.tensor(1.)).如果您尝试传递具有更多值的张量，则会出现错误.

Ok, this works! Now let's try to reproduce error "grad can be implicitly created only for scalar outputs". As you can notice, loss in previous example is a scalar. backward() and grad() by defaults deals with single scalar value: loss.backward(torch.tensor(1.)). If you try to pass tensor with more values you will get an error.

代码:

v = x + 2
y = v ** 2

try:
    dy_hat_dx = grad(outputs=y, inputs=x)
except RuntimeError as err:
    print(err)

输出:

grad 只能为标量输出隐式创建

因此，在使用grad()时需要指定grad_outputs参数如下:

Therefore, when using grad() you need to specify grad_outputs parameter as follows:

代码:

v = x + 2
y = v ** 2

dy_dx = grad(outputs=y, inputs=x, grad_outputs=torch.ones_like(y))
print(f'dy/dx:
 {dy_dx}')

dv_dx = grad(outputs=v, inputs=x, grad_outputs=torch.ones_like(v))
print(f'dv/dx:
 {dv_dx}')

输出:

dy/dx:
(tensor([[6., 6.],[6., 6.]]),)

dv/dx:
(tensor([[1., 1.], [1., 1.]]),)

注意:如果您使用的是 backward()，只需执行 y.backward(torch.ones_like(y)).

NOTE: If you are using backward() instead, simply do y.backward(torch.ones_like(y)).

这篇关于Autograd.grad() 用于 pytorch 中的 Tensor的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！