[functorch] Add l1 loss forward decompositions

functorch/functorch/_src/decompositions.py

@@ -53,17 +53,17 @@ class Reduction(Enum):
 
 
 @register_decomposition(aten.tanh_backward)
-def tanh_backward_decomposition(out_grad: Tensor, y: Tensor):
+def tanh_backward(out_grad: Tensor, y: Tensor):
     return out_grad * (1 - y * y)
 
 
 @register_decomposition(aten.sigmoid_backward)
-def sigmoid_backward_decomposition(out_grad: Tensor, y: Tensor):
+def sigmoid_backward(out_grad: Tensor, y: Tensor):
     return out_grad * (y * (1 - y))
 
 
 @register_decomposition(aten.softplus_backward)
-def softplus_backward_decomposition(out_grad: Tensor, x: Tensor, beta: float, threshold: float):
+def softplus_backward(out_grad: Tensor, x: Tensor, beta: float, threshold: float):
     z = (x * beta).exp()
     return torch.where((x * beta) > threshold, out_grad, out_grad * z / (z + 1.0))
 
@@ -103,7 +103,7 @@ def hardsigmoid(self: Tensor) -> Tensor:
 
 
 @register_decomposition(aten.hardsigmoid_backward)
-def hardsigmoid_backward_decomposition(grad_output: Tensor, self: Tensor):
+def hardsigmoid_backward(grad_output: Tensor, self: Tensor):
     return torch.where((self > -3.0) & (self < 3.0), grad_output * (1.0 / 6.0), grad_output.new_zeros(()))
 
 
@@ -113,7 +113,7 @@ def hardtanh(self: Tensor, min_val: float = -1, max_val: float = 1) -> Tensor:
 
 
 @register_decomposition(aten.hardtanh_backward)
-def hardtanh_backward_decomposition(grad_output: Tensor, self: Tensor, min_val: float, max_val: float):
+def hardtanh_backward(grad_output: Tensor, self: Tensor, min_val: float, max_val: float):
     return torch.where((self <= min_val) | (self >= max_val), grad_output.new_zeros(()), grad_output)
 
 
@@ -133,7 +133,7 @@ def hardswish_backward(grad_output: Tensor, self: Tensor) -> Tensor:
 
 
 @register_decomposition(aten.threshold_backward)
-def threshold_backward_decomposition(grad_output: Tensor, self: Tensor, threshold: float):
+def threshold_backward(grad_output: Tensor, self: Tensor, threshold: float):
     return torch.where(self <= threshold, grad_output.new_zeros((1,)), grad_output)
 
 
@@ -179,7 +179,7 @@ def gelu_backward(grad: Tensor, self: Tensor, approximate: str = 'none'):
 
 
 @register_decomposition(aten.mish_backward)
-def mish_backward_decomposition(grad_output: Tensor, input: Tensor):
+def mish_backward(grad_output: Tensor, input: Tensor):
     input_tanh_softplus = torch.tanh(F.softplus(input))
     input_sigmoid = torch.sigmoid(input)
     out = (input * input_sigmoid * (1 - input_tanh_softplus * input_tanh_softplus))