Address docs for clip_grad functions (pytorch#155125)

This PR takes the opinionated stance that `torch.nn.utils.<func>` should be the preferred API over `torch.nn.utils.clip_grad.<func>`.
Pull Request resolved: pytorch#155125
Approved by: https://github.com/albanD, https://github.com/mikaylagawarecki, https://github.com/janeyx99

Author: jbschlosser

docs/source/conf.py

@@ -1297,10 +1297,6 @@
     "scatter_kwargs",
     # torch.nn.parameter
     "is_lazy",
-    # torch.nn.utils.clip_grad
-    "clip_grad_norm",
-    "clip_grad_norm_",
-    "clip_grad_value_",
     # torch.nn.utils.convert_parameters
     "parameters_to_vector",
     "vector_to_parameters",

torch/nn/utils/clip_grad.py

@@ -16,11 +16,7 @@
 )
 
 
-__all__ = [
-    "clip_grad_norm_",
-    "clip_grad_norm",
-    "clip_grad_value_",
-]
+__all__: list[str] = []
 
 
 _tensor_or_tensors = Union[
@@ -292,3 +288,8 @@ def clip_grad_value_(
         else:
             for grad in grads:
                 cast(Tensor, grad).clamp_(min=-clip_value, max=clip_value)
+
+
+clip_grad_norm.__module__ = "torch.nn.utils"
+clip_grad_norm_.__module__ = "torch.nn.utils"
+clip_grad_value_.__module__ = "torch.nn.utils"