|
No grad & Autocast not working together
|
|
1
|
85
|
October 14, 2025
|
|
How to do back propagation with loss = ||f_{\Theta + \Delta P}(X) - Y||^2 + ||\Delta P||^2 instead of the usual loss = || f_{\Theta + \Delta}(X) - Y ||^2?
|
|
2
|
87
|
October 14, 2025
|
|
"RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [64, 1]], which is output 0 of AsStridedBackward0, is at version 3; expected version 2 instead. Hint: the backtrace further a
|
|
10
|
33482
|
October 9, 2025
|
|
Defining loss that maximizes separation
|
|
0
|
68
|
October 1, 2025
|
|
Ablation Hook I created seems to affect the calculations of the gradients on later layers
|
|
5
|
156
|
September 30, 2025
|
|
How do I set the order of pytorch hooks?
|
|
5
|
145
|
September 29, 2025
|
|
Updating tensors that are used in backpropagation but are not network parameters
|
|
2
|
186
|
September 20, 2025
|
|
Making autograd saved tensors hooks specific to certain arguments
|
|
8
|
410
|
September 17, 2025
|
|
what happens when I use torch.profiler.profile with activities=[torch.profiler.ProfilerActivity.CPU, torch.profiler.ProfilerActivity.CUDA]
|
|
1
|
534
|
September 16, 2025
|
|
Autograd engine thread creation
|
|
2
|
130
|
September 8, 2025
|
|
Solving PDEs using neural networks
|
|
1
|
2502
|
September 5, 2025
|
|
Hessian Vector Product for discounted-return-based loss function
|
|
2
|
115
|
September 2, 2025
|
|
How does torch.where work in autograd
|
|
2
|
129
|
September 2, 2025
|
|
RuntimeError: derivative for aten::grid_sampler_3d_backward is not implemented
|
|
3
|
529
|
August 24, 2025
|
|
2nd derivative for attn is not working
|
|
3
|
728
|
August 20, 2025
|
|
Forward autodiff : Multiplying by python float changes the dual dtype in some situations
|
|
4
|
166
|
August 13, 2025
|
|
Skip the calculation of gradients for some layer after forward pass
|
|
1
|
105
|
August 9, 2025
|
|
PyTorch `torch.no_grad` vs `torch.inference_mode`
|
|
2
|
31752
|
August 5, 2025
|
|
RuntimeError when exponentiating matrix using eigendecomposition
|
|
1
|
94
|
August 5, 2025
|
|
Higher library and create/reatain graph error
|
|
1
|
133
|
August 5, 2025
|
|
When writing a torch.autograd.Function, is there some way for the .backward() to know with respect to which input are the gradients being requested?
|
|
3
|
168
|
July 31, 2025
|
|
Nested forward mode AD not supported (jacobian)— what operations can trigger this?
|
|
1
|
123
|
July 28, 2025
|
|
Verifying grad_x after autograd double backprop
|
|
1
|
107
|
July 24, 2025
|
|
Custom Physics-Informed Loss Function Not Converging for PINN Thermodynamics Model
|
|
0
|
125
|
July 22, 2025
|
|
Can't vmap autograd.grad over outputs
|
|
10
|
386
|
July 16, 2025
|
|
Autograd function returns tensor views
|
|
1
|
91
|
July 16, 2025
|
|
Backward through the graph a second time
|
|
2
|
183
|
July 12, 2025
|
|
LSTM model weight gradients are all zero
|
|
3
|
148
|
July 11, 2025
|
|
Custom loss and its gradient
|
|
1
|
151
|
July 10, 2025
|
|
Why does autograd.backward go one edge further than `inputs`?
|
|
3
|
137
|
July 8, 2025
|