instancenorm: static quant graph mode support (pytorch#39096)

Summary:
Pull Request resolved: pytorch#39096

Hooks up instancenorm for graph mode static quant

Test Plan:
```
python test/test_quantization.py TestQuantizeScriptPTSQOps.test_instance_norm
```

Imported from OSS

Differential Revision: D21885258

fbshipit-source-id: 650cc5b162dda044866176fea6c345082d9788ed

test/quantization/test_quantize_script.py

@@ -2148,6 +2148,26 @@ def test_group_norm(self):
         FileCheck().check_not("aten::group_norm") \
                    .run(m.graph)
 
+    def test_instance_norm(self):
+        data = [(torch.rand((1, 4, 10), dtype=torch.float), torch.randint(0, 1, (1,), dtype=torch.long)) for _ in range(2)]
+        # TODO: handle affine == False (separate PR)
+        instance_norm1d = torch.nn.InstanceNorm1d(4, affine=True)
+        m = self._test_op_impl(instance_norm1d, data, "quantized::instance_norm")
+        FileCheck().check_not("aten::instance_norm") \
+                   .run(m.graph)
+
+        data = [(torch.rand((1, 4, 10, 1), dtype=torch.float), torch.randint(0, 1, (1,), dtype=torch.long)) for _ in range(2)]
+        instance_norm2d = torch.nn.InstanceNorm2d(4, affine=True)
+        m = self._test_op_impl(instance_norm2d, data, "quantized::instance_norm")
+        FileCheck().check_not("aten::instance_norm") \
+                   .run(m.graph)
+
+        data = [(torch.rand((1, 4, 10, 1, 1), dtype=torch.float), torch.randint(0, 1, (1,), dtype=torch.long)) for _ in range(2)]
+        instance_norm3d = torch.nn.InstanceNorm3d(4, affine=True)
+        m = self._test_op_impl(instance_norm3d, data, "quantized::instance_norm")
+        FileCheck().check_not("aten::instance_norm") \
+                   .run(m.graph)
+
     def test_quantize_general_shape_ops(self):
         """ A test that checks dequantize will be swapped for
         all supported general shape ops like aten::flatten

torch/csrc/jit/passes/quantization/helper.cpp

@@ -22,6 +22,7 @@ std::vector<std::string> _static_quantizable_call_funcs = {
     "hardswish",
     "layer_norm",
     "group_norm",
+    "instance_norm",
 };
 
 std::vector<std::string> _static_quantizable_aten_funcs = {
@@ -34,6 +35,7 @@ std::vector<std::string> _static_quantizable_aten_funcs = {
     "hardswish",
     "layer_norm",
     "group_norm",
+    "instance_norm",
 };
 
 std::vector<std::string> _dynamic_quantizable_call_funcs = {

torch/csrc/jit/passes/quantization/quantization_patterns.h

@@ -620,6 +620,19 @@ graph(%a_quant, %num_groups, %weight, %bias, %eps, %cudnn_enabled, %output_scale
          %r = quantized::group_norm(%a_quant, %num_groups, %weight, %bias, %eps, %output_scale, %output_zero_point)
          return (%r) )";
 
+  // quantized::instance_norm
+  std::string instance_norm = R"(
+graph(%a_quant, %weight, %bias, %running_mean, %running_var, %use_input_stats, %momentum, %eps, %cudnn_enabled, %output_scale, %output_zero_point, %scalar_type):
+         %a_dequant = aten::dequantize(%a_quant)
+         %r_in = aten::instance_norm(%a_dequant, %weight, %bias, %running_mean, %running_var, %use_input_stats, %momentum, %eps, %cudnn_enabled)
+         %r = aten::quantize_per_tensor(%r_in, %output_scale, %output_zero_point, %scalar_type)
+         return (%r) )";
+
+  std::string quantized_instance_norm = R"(
+graph(%a_quant, %weight, %bias, %running_mean, %running_var, %use_input_stats, %momentum, %eps, %cudnn_enabled, %output_scale, %output_zero_point, %scalar_type):
+         %r = quantized::instance_norm(%a_quant, %weight, %bias, %eps, %output_scale, %output_zero_point)
+         return (%r) )";
+
   // ============= General Ops that inherit quantization paramters from input
   // tensor =============
   auto avg_pool1d = getInputTensorQParamOpFusionInfo(
@@ -820,6 +833,7 @@ graph(%a_quant, %num_groups, %weight, %bias, %eps, %cudnn_enabled, %output_scale
       {"quantized::hardswish", hardswish, quantized_hardswish},
       {"quantized::layer_norm", layer_norm, quantized_layer_norm},
       {"quantized::group_norm", group_norm, quantized_group_norm},
+      {"quantized::instance_norm", instance_norm, quantized_instance_norm},
       avg_pool1d,
       avg_pool2d,
       avg_pool3d,