|
About the autograd category
|
|
0
|
2421
|
May 13, 2017
|
|
Performance gap between `batch_size==32` and `batch_size==8, gradient_accumulation==4`
|
|
1
|
25
|
September 25, 2022
|
|
How backward works for two dependent forward pass
|
|
3
|
32
|
September 25, 2022
|
|
Gradient computation failed with torch.stack
|
|
5
|
356
|
September 25, 2022
|
|
What are the parameters involved to calculate the gradient during backpropagation?
|
|
1
|
22
|
September 25, 2022
|
|
Manual application of the chain rule
|
|
2
|
37
|
September 24, 2022
|
|
Calculating the gradient of a sparse Tensor argument to mm is not supported
|
|
1
|
596
|
September 24, 2022
|
|
Why is `retain_graph=True` needed in some case but not in other
|
|
11
|
102
|
September 23, 2022
|
|
Extending Pytorch Gradients
|
|
0
|
25
|
September 23, 2022
|
|
What does this parameter mean?(torch.ones_like(inp)
|
|
1
|
24
|
September 23, 2022
|
|
Cant' find : RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation
|
|
1
|
41
|
September 22, 2022
|
|
Grad is None even when requires_grad=True
|
|
9
|
27075
|
September 22, 2022
|
|
How to compute the element-wise gradients efficiently?
|
|
3
|
250
|
September 22, 2022
|
|
Is there a simple way to make the updated paramaters as interim tensors?
|
|
0
|
29
|
September 22, 2022
|
|
DataParallel/LOSS element 0 of tensors does not require grad and does not have a grad_fn
|
|
2
|
34
|
September 22, 2022
|
|
Sudden explosion of the training loss
|
|
0
|
35
|
September 22, 2022
|
|
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation (Meta Learning)
|
|
8
|
64
|
September 21, 2022
|
|
Gradients w.r.t. loss component
|
|
2
|
49
|
September 21, 2022
|
|
Element-wise multiplication for model parameters and self-defined mask in traing workflow
|
|
3
|
47
|
September 21, 2022
|
|
Is save_on_cpu() overlaps the computation and communication?
|
|
3
|
54
|
September 20, 2022
|
|
Error when performing autograd gradcheck twice on torchscript
|
|
3
|
50
|
September 20, 2022
|
|
Cannot find where's the in-place operation error
|
|
5
|
235
|
September 20, 2022
|
|
Strange CUDA error when backward
|
|
2
|
397
|
September 20, 2022
|
|
Getting correct prediction for only one class
|
|
2
|
39
|
September 19, 2022
|
|
Keep the gradient function when changing the data type of tensor
|
|
1
|
45
|
September 18, 2022
|
|
Training two models sequentially, having different batch sizes
|
|
5
|
81
|
September 18, 2022
|
|
Custom Backward Pass for Trilinear Interpolation
|
|
0
|
48
|
September 17, 2022
|
|
Not getting symmetric hessian when calculating by using torch.autograd
|
|
5
|
66
|
September 17, 2022
|
|
Understanding autograd memory allocation during forward pass
|
|
0
|
35
|
September 16, 2022
|
|
Only 1 loss and multi-branch training
|
|
0
|
35
|
September 16, 2022
|