本文是通过学习 AI研习社 炼丹兄 所做笔记！！！
1.1 .grad
例：，计算y关于w的梯度：
(上式计算中，)
import torch
w = torch.tensor([1.],requires_grad = True)
x = torch.tensor([2.],requires_grad = True)a = w + x
b = w + 1
y = a * by.backward()
print(w.grad)1.2 .is_leaf
例：，计算y关于w的梯度：
(上式计算中，)
import torch
w = torch.tensor([1.],requires_grad = True)
x = torch.tensor([2.],requires_grad = True)a = w + x
b = w + 1
y = a * by.backward()
print(w.is_leaf,x.is_leaf,a.is_leaf,b.is_leaf,y.is_leaf)
print(w.grad,x.grad,a.grad,b.grad,y.grad) 1.3 .retain_grad()
例：，计算y关于w的梯度：
(上式计算中，)
import torch
w = torch.tensor([1.],requires_grad = True)
x = torch.tensor([2.],requires_grad = True)a = w + x
a.retain_grad()
b = w + 1
y = a * by.backward()
print(w.is_leaf,x.is_leaf,a.is_leaf,b.is_leaf,y.is_leaf)
print(w.grad,x.grad,a.grad,b.grad,y.grad)1.4 .grad_fn
例：，计算y关于w的梯度：
(上式计算中，)
import torchw = torch.tensor([1.],requires_grad = True)
x = torch.tensor([2.],requires_grad = True)a = w + x
a.retain_grad()
b = w + 1
y = a * by.backward()
print(y.grad_fn)
print(a.grad_fn)
print(w.grad_fn)本文链接：https://my.lmcjl.com/post/11265.html
展开阅读全文
4 评论