Add support for quantized LeakyReLU

Summary: Also adds support for backend_config Reviewed By: mcr229 Differential Revision: D47043207 fbshipit-source-id: 509bd4c02eb7ff5d3d47762522debd827bee7240

backends/xnnpack/partition/xnnpack_partitioner.py

@@ -553,6 +553,9 @@ def __init__(self):
  torch.nn.ReLU,
  torch.nn.functional.relu,
  torch.nn.functional.relu_,
+ torch.nn.functional.leaky_relu,
+ torch.nn.functional.leaky_relu_,
+ torch.nn.LeakyReLU,
 ]
 
 # Modules which support dynamic quantization

backends/xnnpack/test/test_xnnpack_quantized.py

@@ -178,6 +178,47 @@ def test_xnnpack_qhardtanh(self):
  example_inputs = (torch.randn(1, 1, 1),)
  self.quantize_and_test_model(torch.nn.Hardtanh(), example_inputs)
 
+ def test_xnnpack_leaky_relu(self):
+ example_inputs = (torch.randn(1, 3, 3),)
+
+ class LeakyReLUModule(torch.nn.Module):
+ def __init__(self):
+ super().__init__()
+ self.leaky_relu_out_of_place = torch.nn.LeakyReLU(negative_slope=0.2)
+
+ def forward(self, x):
+ return self.leaky_relu_out_of_place(x)
+
+ self.quantize_and_test_model(LeakyReLUModule(), example_inputs)
+
+ def test_xnnpack_leaky_relu2(self):
+ example_inputs = (torch.randn(1, 3, 3),)
+
+ class LeakyReLUModule(torch.nn.Module):
+ def __init__(self):
+ super().__init__()
+ self.leaky_relu_in_place = torch.nn.LeakyReLU(
+ negative_slope=0.08, inplace=True
+ )
+
+ def forward(self, x):
+ return self.leaky_relu_in_place(x)
+
+ self.quantize_and_test_model(LeakyReLUModule(), example_inputs)
+
+ def test_xnnpack_leaky_relu3(self):
+ example_inputs = (torch.randn(1, 3, 3),)
+
+ class LeakyReLUModule(torch.nn.Module):
+ def __init__(self):
+ super().__init__()
+ self.leaky_relu_functional_default = torch.nn.functional.leaky_relu
+
+ def forward(self, x):
+ return self.leaky_relu_functional_default(x)
+
+ self.quantize_and_test_model(LeakyReLUModule(), example_inputs)
+
  def test_xnnpack_qlinear(self):
  in_size = 1
  input_size = 3