Backprop trough Fully connected network and sparse matrix (1)
[SOLVED] Loss.backward hangs (6)
Confusing code on RuntimeError: Trying to backward through the graph a second time, but the buffers have already been freed. Specify retain_graph=True when calling backward the first time (4)
Different Losses on 2 different machines ( 2 ) (27)
Effective computation of single gradients for minibatches (5)
Define back-prop on a custom layer by enforcing the sum of its weight to equal 1 (2)
How to change the gradients of some parameters? (1)
Manually changing the gradients of parameters (3)
Define a weighted linear combination of matrices layer (1)
Selectively computing gradients instead of the whole network (3)
Some problems in custom loss functions and so on (3)
Model parallelism in Multi-GPUs: forward/backward graph (14)
Parameter Lists in Pytorch (4)
Learn initial hidden state (h0) for RNN (13)
Efficient detached vector Jacobian product (1)
Convert tuple to tensor without breaking graph (9)
Gradcheck works on functions indepedently, but not in sequence (3)
Shape of the validation loss (1)
Fail to re-create computational graph in a for-loop (4)
Keep running in to "Trying to backward through the graph a second time" Error as a beginner (4)
Kernel size can't greater than actual input size (4)
Only update weights in regard to one of the outputs (5)
Autograd raise an error : NaNs encountered when trying to perform matrix-vector multiplication (1)
Inference when output sigmoid is within BCEWithLogitsLoss (4)
Prioritize Gradients from Different Loss Functions (2)
Saving a Variable with requires_grad=True in a dictionary is not being updated ( 2 ) (21)
Do ReLU and Dropout layers take up space on backward pass (1)
Weird Behavior in CrossEntropyLoss (1)
Calculating the Hessian vector product with nn.parameters() (1)
[solved] Getting batches of gradients (4)