torch.optim.functional#
Created On: Aug 27, 2026 | Last Updated On: Aug 27, 2026
Functional optimizer APIs give callers complete control over the parameters, gradients, and optimizer state. They are not pure functions; they update parameters and optimizer-state tensors in place. See Functional optimizer API for their common contract and usage examples.
Functional optimizer APIs.
adadelta |
Functional API that performs Adadelta algorithm computation. |
adafactor |
Functional API that performs Adafactor algorithm computation. |
adagrad |
Functional API that performs Adagrad algorithm computation. |
adam |
Functional API that performs Adam algorithm computation. |
adamax |
Functional API that performs Adamax algorithm computation. |
adamw |
Functional API that performs AdamW algorithm computation. |
asgd |
Functional API that performs ASGD algorithm computation. |
muon |
Functional API that performs Muon algorithm computation. |
nadam |
Functional API that performs NAdam algorithm computation. |
radam |
Functional API that performs RAdam algorithm computation. |
rmsprop |
Functional API that performs RMSprop algorithm computation. |
rprop |
Functional API that performs Rprop algorithm computation. |
sgd |
Functional API that performs SGD algorithm computation. |
sparse_adam |
Functional API that performs SparseAdam algorithm computation. |