|
About the autograd category
|
|
0
|
1762
|
May 13, 2017
|
|
Can I do modification on a branch of chain in the backward pass?
|
|
3
|
25
|
October 23, 2021
|
|
Multi-task multi-loss learning
|
|
3
|
42
|
October 22, 2021
|
|
Custom activation functions?
|
|
7
|
2156
|
October 22, 2021
|
|
Autograd fails without giving any warning while doing matrix operations
|
|
10
|
55
|
October 22, 2021
|
|
Using input tensors for loss instead of output y_pred
|
|
11
|
56
|
October 22, 2021
|
|
Checkpoint breaks grads
|
|
1
|
24
|
October 21, 2021
|
|
"requires_grad" after list addition
|
|
1
|
25
|
October 21, 2021
|
|
Why did you refactor autograd.Function by using static functions?
|
|
1
|
31
|
October 20, 2021
|
|
Can the new functional autograd take batches? Also, is it more efficient to compute a hessian with the new functional autograd than it is using the old autograd?
|
|
15
|
487
|
October 20, 2021
|
|
.to(device) slowing the code
|
|
5
|
389
|
October 20, 2021
|
|
Help!Why grad is always NA
|
|
1
|
18
|
October 19, 2021
|
|
Disable "in-place" updates in troch.nn
|
|
4
|
49
|
October 18, 2021
|
|
RuntimeError: element 0 of variables does not require grad and does not have a grad_fn
|
|
49
|
95707
|
October 18, 2021
|
|
Model partial derivatives into loss function
|
|
2
|
26
|
October 18, 2021
|
|
Is it possible to get the Hessian of the loss with respect to the output of a layer via `register_full_backward_hook`
|
|
9
|
95
|
October 18, 2021
|
|
Derivation of output w.r.t input features returns 'None' after setting 'allow_unused=True'
|
|
4
|
47
|
October 18, 2021
|
|
Calling loss.backward() multiple times vs. aggregating losses
|
|
1
|
31
|
October 17, 2021
|
|
Learning rate new object difference
|
|
1
|
29
|
October 16, 2021
|
|
Pytorch + pytest + multigpu
|
|
0
|
22
|
October 16, 2021
|
|
What's the most efficient way to get the Hessian of the loss with respect to all parameters for all input samples?
|
|
0
|
25
|
October 16, 2021
|
|
Is it possible to hijack the storage of data needed in backward pass
|
|
2
|
35
|
October 15, 2021
|
|
Efficient way to divide samples' gradient contribution
|
|
0
|
28
|
October 15, 2021
|
|
RuntimeError: set_sizes_and_strides is not allowed on a Tensor created from .data or .detach(). when using Gradient checkpointing
|
|
1
|
30
|
October 15, 2021
|
|
Out of memory issue on both cpu and gpu
|
|
1
|
34
|
October 14, 2021
|
|
PyTorch Profiler kineto_event.name() shows "UnicodeDecodeError: 'utf-8' codec can't decode byte 0x89 in position 5: invalid start byte"
|
|
1
|
78
|
October 14, 2021
|
|
Consistent Memory Leak Project Gradient Descent
|
|
0
|
34
|
October 13, 2021
|
|
Custom loss function leading to loss of gradients
|
|
4
|
73
|
October 13, 2021
|
|
Multi-dimensional sequence to sequence network: how to ignore inplace operation?
|
|
0
|
17
|
October 13, 2021
|
|
Grad value when copying tensor
|
|
2
|
44
|
October 13, 2021
|
