Function torch::autograd::grad¶
Defined in File autograd.h
Function Documentation¶
-
variable_list
torch::autograd
::
grad
(const variable_list &outputs, const variable_list &inputs, const variable_list &grad_outputs = {}, c10::optional<bool> retain_graph = c10::nullopt, bool create_graph = false, bool allow_unused = false)¶ Computes and returns the sum of gradients of outputs with respect to the inputs.
grad_outputs
should be a sequence of length matchingoutput
containing the “vector” in Jacobian-vector product, usually the pre-computed gradients w.r.t. each of the outputs. If an output doesn’t require_grad, then the gradient can betorch::Tensor()
).- Parameters
outputs
: outputs of the differentiated function.inputs
: Inputs w.r.t. which the gradient will be returned (and not accumulated intoat::Tensor::grad
).grad_outputs
: The “vector” in the Jacobian-vector product. Usually gradients w.r.t. each output.torch::Tensor()
values can be specified for scalar Tensors or ones that don’t require grad. If atorch::Tensor()
value would be acceptable for all grad_tensors, then this argument is optional. Default:{}
.retain_graph
: Iffalse
, the graph used to compute the grad will be freed. Note that in nearly all cases setting this option totrue
is not needed and often can be worked around in a much more efficient way. Defaults to the value ofcreate_graph
.create_graph
: Iftrue
, graph of the derivative will be constructed, allowing to compute higher order derivative products. Default:false
.allow_unused
: Iffalse
, specifying inputs that were not used when computing outputs (and therefore their grad is always zero) is an error. Defaults tofalse
.