pytorch : grad can be implicitly created only for scalar outputs

错误信息

  File "***.py", line 101, in trainloss.backward()File "***/anaconda3/envs/ngepc/lib/python3.8/site-packages/torch/tensor.py", line 221, in backwardtorch.autograd.backward(self, gradient, retain_graph, create_graph)File "***/anaconda3/envs/ngepc/lib/python3.8/site-packages/torch/autograd/__init__.py", line 126, in backwardgrad_tensors_ = _make_grads(tensors, grad_tensors_)File "***/anaconda3/envs/ngepc/lib/python3.8/site-packages/torch/autograd/__init__.py", line 50, in _make_gradsraise RuntimeError("grad can be implicitly created only for scalar outputs")
RuntimeError: grad can be implicitly created only for scalar outputs

原因是函数autograd.grad()的grad_outputs 参数默认是output为数值，Mse损失参数’none’: no reduction will be applied，将输出矩阵，需要输出矩阵元素的和或均值已得到数值而非矩阵的loss,所以修改如下：

将

self.l = nn.MSELoss(reduction='nobe')

改为：

self.l = nn.MSELoss(reduction='mean')

类似于：将

 a=Variable(torch.FloatTensor([1,2,3]),requires_grad=True)autograd.grad(outputs=a,inputs=a)

改为

 a=Variable(torch.FloatTensor([1,2,3]),requires_grad=True)autograd.grad(outputs=(a.sum()/a.shape[1]),inputs=a)

pytorch : grad can be implicitly created only for scalar outputs相关推荐

pytorch: grad can be implicitly created only for scalar outputs 的解决办法
如果X.grad报错: grad can be implicitly created only for scalar outputs 加上去 X.backward(torch.ones_like(X) ...
pytorch 1.9.0 backward函数解释以及报错（RuntimeError: grad can be implicitly created only for scalar outputs）
文章目录官方文档简单示例示例1 示例2(报错(RuntimeError: grad can be implicitly created only for scalar outputs)解决方法) ...
grad can be implicitly created only for scalar outputs
1.Autograd:自动求导 torch.Tensor 是这个包的核心类.如果设置它的属性 .requires_grad 为 True,那么它将会追踪对于该张量的所有操作.当完成计算后可以通过调 ...
DEBUG:grad can be implicitly created only for scalar outputs
DEBUG:grad can be implicitly created only for scalar outputs 解决: 添加 retain_graph=True 只能起效一次
(已解决)多卡训练时报错RuntimeError: grad can be implicitly created only for scalar outputs
背景博主第一次使用多卡训练,在程序中添加了如下代码 # 包装为并行风格模型 os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID& ...
RuntimeError: grad can be implicitly created only for scalar outputs的原因：Pytorch不支持对张量的求导
一.背景介绍原则上,Pytorch不支持对张量的求导,即如果z是张量的话,需要先将其转为标量. 浏览了很多博客,给出的解决方案都是说在求导时,加一个torch.ones_like(z)的参数. 下面 ...
RuntimeError: grad can be implicitly created only for scalar outputs
这个的原因是梯度只能为标量(即一个数)输出隐式地创建也就是说传递的时候loss.backward()需要加上一个参数 loss.backward(torch.ones_like(loss)) 这样就 ...
高效理解pytorch的backward需要scalar outputs
利用backward时 , 可能经常遇到错误 RuntimeError: grad can be implicitly created only for scalar outputs 理解的最好方式就 ...
PyTorch grad 与 Optimizer(params) 区别
目录 PyTorch grad 与 Optimizer(params) 区别 PyTorch grad 与 Optimizer(params) 区别 Tensor 可以设置属性 requires_gr ...

pytorch : grad can be implicitly created only for scalar outputs

pytorch : grad can be implicitly created only for scalar outputs相关推荐

最新文章

热门文章