chore: add additional BN native converter (#2446)

Signed-off-by: Dheeraj Peri <[email protected]>
pytorch · Nov 9, 2023 · bfa1bdf · bfa1bdf
1 parent ae733c6
commit bfa1bdf
Show file tree

Hide file tree

Showing 3 changed files with 50 additions and 1 deletion.
diff --git a/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py b/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py
@@ -89,6 +89,37 @@ def aten_ops_batch_norm(
     )
 
 
+@dynamo_tensorrt_converter(
+    torch.ops.aten._native_batch_norm_legit_no_training.default,
+    capability_validator=one_user_validator,
+)
+def aten_ops_batch_norm_legit_no_training(
+    ctx: ConversionContext,
+    target: Target,
+    args: Tuple[Argument, ...],
+    kwargs: Dict[str, Argument],
+    name: str,
+) -> Union[TRTTensor, Sequence[TRTTensor]]:
+    return impl.normalization.batch_norm(
+        ctx,
+        target,
+        SourceIR.ATEN,
+        name,
+        input=args[0],
+        weight=args[1],
+        bias=args[2],
+        running_mean=args[3],
+        running_var=args[4],
+        training=False,
+        momentum=args[5],
+        eps=args[6],
+        cudnn_enabled=False,
+        return_mean_rstd=(
+            target == torch.ops.aten._native_batch_norm_legit_no_training.default
+        ),
+    )
+
+
 @dynamo_tensorrt_converter(
     torch.ops.aten.native_layer_norm.default, capability_validator=one_user_validator
 )

diff --git a/py/torch_tensorrt/dynamo/lowering/_decomposition_groups.py b/py/torch_tensorrt/dynamo/lowering/_decomposition_groups.py
@@ -100,7 +100,6 @@
     aten.native_batch_norm_backward,
     aten._native_batch_norm_legit,
     aten._native_batch_norm_legit_functional,
-    aten._native_batch_norm_legit_no_training,
     aten.native_dropout_backward,
     aten.native_group_norm_backward,
     aten.native_layer_norm_backward,

diff --git a/tests/py/dynamo/conversion/test_batch_norm_aten.py b/tests/py/dynamo/conversion/test_batch_norm_aten.py
@@ -107,6 +107,25 @@ def forward(self, x):
             inputs,
         )
 
+    def test_batchnorm_legit_no_training(self):
+        class BatchNorm(torch.nn.Module):
+            def forward(self, x):
+                return torch.ops.aten._native_batch_norm_legit_no_training.default(
+                    x,
+                    torch.ones((FEATURE_NUM,)),
+                    torch.zeros((FEATURE_NUM,)),
+                    torch.zeros((FEATURE_NUM,)),
+                    torch.ones((FEATURE_NUM,)),
+                    0.1,
+                    1e-05,
+                )[0]
+
+        inputs = [torch.randn(1, 3, 224, 224)]
+        self.run_test(
+            BatchNorm(),
+            inputs,
+        )
+
     def test_batchnorm1d_with_dynamic_shape(self):
         class BatchNorm(torch.nn.Module):
             def forward(self, x):