'ExponentialLR' object has no attribute 'zero_grad'

The error message you are seeing suggests that you are trying to call the zero_grad() method on an instance of the ExponentialLR class, which does not have this method.
zero_grad() is a method defined in the torch.optim.Optimizer class, which is used to set the gradients of all the parameters to zero before backpropagation.
You should make sure that you are calling zero_grad() on an instance of the Optimizer class, not the ExponentialLR class. Here is an example of how to use zero_grad() properly:
import torch
import torch.optim as optim

optimizer = optim.SGD(model.parameters(), lr=0.1)
scheduler = optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.9)

# perform a forward pass and compute loss
loss.backward()

# set gradients to zero
optimizer.zero_grad()

# perform backpropagation
optimizer.step()

# update learning rate
scheduler.step()
登录后复制
In this example, zero_grad() is called on the optimizer instance, not the scheduler instance.