fix: Implement aten.mean.default and aten.mean.dim converters (#1810)

Author: gs-olive

py/torch_tensorrt/fx/converters/aten_ops_converters.py

@@ -41,7 +41,6 @@ def aten_ops_add(
     return acc_ops_converters.acc_ops_add(network, target, None, kwargs_new, name)
 
 
-@tensorrt_converter(torch.ops.aten.mean.dim)
 @tensorrt_converter(torch.ops.aten._adaptive_avg_pool3d.default)
 @tensorrt_converter(torch.ops.aten._adaptive_avg_pool2d.default)
 def aten_ops_adaptive_avg_poolnd(
@@ -51,24 +50,38 @@ def aten_ops_adaptive_avg_poolnd(
     kwargs: Dict[str, Argument],
     name: str,
 ) -> Union[TRTTensor, Sequence[TRTTensor]]:
-    if target == torch.ops.aten.mean.dim:
-
-        if list(args[1]) != [-1, -2]:
-            raise RuntimeError(f"We do not support {target} has dim={args[1]}")
-        else:
-            output_size = [1, 1]
-    else:
-        output_size = args[1]
-
     kwargs_new = {
         "input": args[0],
-        "output_size": output_size,
+        "output_size": args[1],
     }
     return acc_ops_converters.acc_ops_adaptive_avg_poolnd(
         network, target, None, kwargs_new, name
     )
 
 
+@tensorrt_converter(torch.ops.aten.mean.default)
+@tensorrt_converter(torch.ops.aten.mean.dim)
+def aten_ops_mean(
+    network: TRTNetwork,
+    target: Target,
+    args: Tuple[Argument, ...],
+    kwargs: Dict[str, Argument],
+    name: str,
+) -> TRTTensor:
+    # Default invocation of aten.mean only uses first argument and
+    # averages over all elements (all dimensions)
+    # aten.mean.dim invocation allows specification of dimensions to average
+    # over, as well at the option to keep the dimension or not
+    kwargs_new = {
+        "input": args[0],
+        "dim": args[1] if len(args) >= 2 else list(range(len(args[0].shape))),
+        "keepdim": args[2] if len(args) >= 3 else False,
+    }
+    return add_reduce_layer(
+        network, target, args, kwargs_new, trt.ReduceOperation.AVG, name
+    )
+
+
 @tensorrt_converter(torch.ops.aten.batch_norm)
 def aten_ops_batch_norm(
     network: TRTNetwork,

py/torch_tensorrt/fx/test/converters/aten_op/test_mean_aten.py

@@ -0,0 +1,84 @@
+import torch
+import torch.nn as nn
+from torch.testing._internal.common_utils import run_tests
+from torch_tensorrt.fx.tools.common_fx2trt import DispatchTestCase, InputTensorSpec
+
+
+class TestMeanDimConverter(DispatchTestCase):
+    def test_mean_dim_keepdims(self):
+        class TestModule(nn.Module):
+            def forward(self, x):
+                return torch.mean(x, dim=[0, 1], keepdim=True)
+
+        inputs = [torch.randn(1, 10)]
+        self.run_test(TestModule(), inputs, expected_ops={torch.ops.aten.mean.dim})
+
+    def test_mean_dim_keepdims_with_dynamic_shape(self):
+        class TestModule(nn.Module):
+            def forward(self, x):
+                return torch.mean(x, dim=[0, 1, 2], keepdim=True)
+
+        input_specs = [
+            InputTensorSpec(
+                shape=(-1, -1, -1),
+                dtype=torch.float32,
+                shape_ranges=[((1, 1, 1), (1, 2, 3), (3, 3, 3))],
+            ),
+        ]
+        self.run_test_with_dynamic_shape(
+            TestModule(), input_specs, expected_ops={torch.ops.aten.mean.dim}
+        )
+
+    def test_mean_dim_keepdims_false(self):
+        class TestModule(nn.Module):
+            def forward(self, x):
+                return torch.mean(x, dim=0, keepdim=False)
+
+        inputs = [torch.randn(3, 5, 7)]
+        self.run_test(TestModule(), inputs, expected_ops={torch.ops.aten.mean.dim})
+
+    def test_mean_dim_keepdims_false_with_dynamic_shape(self):
+        class TestModule(nn.Module):
+            def forward(self, x):
+                return torch.mean(x, dim=-1, keepdim=False)
+
+        input_specs = [
+            InputTensorSpec(
+                shape=(-1, -1, -1),
+                dtype=torch.float32,
+                shape_ranges=[((1, 1, 1), (1, 2, 3), (3, 3, 3))],
+            ),
+        ]
+        self.run_test_with_dynamic_shape(
+            TestModule(), input_specs, expected_ops={torch.ops.aten.mean.dim}
+        )
+
+
+class TestMeanConverter(DispatchTestCase):
+    def test_mean(self):
+        class TestModule(nn.Module):
+            def forward(self, x):
+                return torch.mean(x)
+
+        inputs = [torch.randn(3, 8, 5, 7, 1)]
+        self.run_test(TestModule(), inputs, expected_ops={torch.ops.aten.mean.default})
+
+    def test_mean_with_dynamic_shape(self):
+        class TestModule(nn.Module):
+            def forward(self, x):
+                return torch.mean(x)
+
+        input_specs = [
+            InputTensorSpec(
+                shape=(-1, -1, -1),
+                dtype=torch.float32,
+                shape_ranges=[((1, 1, 1), (1, 5, 8), (3, 10, 10))],
+            ),
+        ]
+        self.run_test_with_dynamic_shape(
+            TestModule(), input_specs, expected_ops={torch.ops.aten.mean.default}
+        )
+
+
+if __name__ == "__main__":
+    run_tests()