How does one check if a tensor/parameter in pytorch is trainable?
1 Like
check if it has flag requires_grad to true:
Python 3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56)
[GCC 7.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> x = torch.randn(3,4)
>>> x
tensor([[ 0.5905, 0.7622, 0.5487, 0.4738],
[-0.2488, 0.1633, 2.3185, -0.6563],
[ 1.0734, -0.3375, -0.5587, 0.3377]])
>>> x.grad
>>> x.requires_grad
False
>>> x.requires_grad = True
>>> x
tensor([[ 0.5905, 0.7622, 0.5487, 0.4738],
[-0.2488, 0.1633, 2.3185, -0.6563],
[ 1.0734, -0.3375, -0.5587, 0.3377]], requires_grad=True)
>>>
source: https://pytorch.org/tutorials/beginner/blitz/autograd_tutorial.html
1 Like
Yes a Tensor will have it’s .grad field populated if requires_grad=True and is_leaf=True.
1 Like