functorch.grad¶

functorch.
grad
(func, argnums=0, has_aux=False)[source]¶ grad
operator helps computing gradients offunc
with respect to the input(s) specified byargnums
. This operator can be nested to compute higherorder gradients. Parameters
func (Callable) – A Python function that takes one or more arguments. Must return a singleelement Tensor. If specified
has_aux
equalsTrue
, function can return a tuple of singleelement Tensor and other auxiliary objects:(output, aux)
.argnums (int or Tuple[int]) – Specifies arguments to compute gradients with respect to.
argnums
can be single integer or tuple of integers. Default: 0.has_aux (bool) – Flag indicating that
func
returns a tensor and other auxiliary objects:(output, aux)
. Default: False.
 Returns
Function to compute gradients with respect to its inputs. By default, the output of the function is the gradient tensor(s) with respect to the first argument. If specified
has_aux
equalsTrue
, tuple of gradients and output auxiliary objects is returned. Ifargnums
is a tuple of integers, a tuple of output gradients with respect to eachargnums
value is returned.
Example of using
grad
:>>> # xdoctest: +SKIP >>> from torch.func import grad >>> x = torch.randn([]) >>> cos_x = grad(lambda x: torch.sin(x))(x) >>> assert torch.allclose(cos_x, x.cos()) >>> >>> # Secondorder gradients >>> neg_sin_x = grad(grad(lambda x: torch.sin(x)))(x) >>> assert torch.allclose(neg_sin_x, x.sin())
When composed with
vmap
,grad
can be used to compute persamplegradients:>>> # xdoctest: +SKIP >>> from torch.func import grad, vmap >>> batch_size, feature_size = 3, 5 >>> >>> def model(weights, feature_vec): >>> # Very simple linear model with activation >>> assert feature_vec.dim() == 1 >>> return feature_vec.dot(weights).relu() >>> >>> def compute_loss(weights, example, target): >>> y = model(weights, example) >>> return ((y  target) ** 2).mean() # MSELoss >>> >>> weights = torch.randn(feature_size, requires_grad=True) >>> examples = torch.randn(batch_size, feature_size) >>> targets = torch.randn(batch_size) >>> inputs = (weights, examples, targets) >>> grad_weight_per_example = vmap(grad(compute_loss), in_dims=(None, 0, 0))(*inputs)
Example of using
grad
withhas_aux
andargnums
:>>> # xdoctest: +SKIP >>> from torch.func import grad >>> def my_loss_func(y, y_pred): >>> loss_per_sample = (0.5 * y_pred  y) ** 2 >>> loss = loss_per_sample.mean() >>> return loss, (y_pred, loss_per_sample) >>> >>> fn = grad(my_loss_func, argnums=(0, 1), has_aux=True) >>> y_true = torch.rand(4) >>> y_preds = torch.rand(4, requires_grad=True) >>> out = fn(y_true, y_preds) >>> # > output is ((grads w.r.t y_true, grads w.r.t y_preds), (y_pred, loss_per_sample))
Note
Using PyTorch
torch.no_grad
together withgrad
.Case 1: Using
torch.no_grad
inside a function:>>> # xdoctest: +SKIP >>> def f(x): >>> with torch.no_grad(): >>> c = x ** 2 >>> return x  c
In this case,
grad(f)(x)
will respect the innertorch.no_grad
.Case 2: Using
grad
insidetorch.no_grad
context manager:>>> # xdoctest: +SKIP >>> with torch.no_grad(): >>> grad(f)(x)
In this case,
grad
will respect the innertorch.no_grad
, but not the outer one. This is becausegrad
is a “function transform”: its result should not depend on the result of a context manager outside off
.Warning
We’ve integrated functorch into PyTorch. As the final step of the integration, functorch.grad is deprecated as of PyTorch 2.0 and will be deleted in a future version of PyTorch >= 2.3. Please use torch.func.grad instead; see the PyTorch 2.0 release notes and/or the torch.func migration guide for more details https://pytorch.org/docs/master/func.migrating.html