torch.optim.functional.sgd

torch.optim.functional.sgd(params, d_p_list, momentum_buffer_list, has_sparse_grad=False, foreach=None, fused=None, grad_scale=None, found_inf=None, *, weight_decay, momentum, lr, dampening, nesterov, maximize)

Functional API that performs SGD algorithm computation.

This function updates the provided parameters and optimizer state in place. The caller must initialize and retain optimizer state. Unless intentionally constructing a differentiable update with a supported differentiable=True argument, call this function under torch.no_grad. See Functional optimizer API for the common functional optimizer contract and examples, and SGD for algorithm details.