[Relax][PyTorch] Add Adaptive AvgPool 1D and 3D Op Support for Exported Program and FX graph

python/tvm/relax/frontend/torch/base_fx_graph_translator.py

@@ -449,12 +449,53 @@ def _isin(self, node: fx.Node) -> relax.Var:
 
     ########## Neural Network ##########
 
+    def _adaptive_avg_pool1d(self, node: fx.Node) -> relax.Var:
+        x = self.env[node.args[0]]
+        output_size = node.args[1] if len(node.args) > 1 else node.kwargs["output_size"]
+        # Expand to 3D by adding batch dim if input is 2D
+        x_ndim = x.struct_info.ndim
+        if x_ndim == 2:
+            x = relax.op.expand_dims(x, axis=0)
+
+        result = self.block_builder.emit(
+            relax.op.nn.adaptive_avg_pool1d(x, output_size, layout="NCW")
+        )
+        # Remove added batch dim from result
+        if x_ndim == 2:
+            result = relax.op.squeeze(result, axis=[0])
+        return result
+
     def _adaptive_avg_pool2d(self, node: fx.Node) -> relax.Var:
         x = self.env[node.args[0]]
         output_size = node.args[1]
-        return self.block_builder.emit(
+        # Expand to 4D by adding batch dim if input is 3D
+        x_ndim = x.struct_info.ndim
+        if x_ndim == 3:
+            x = relax.op.expand_dims(x, axis=0)
+
+        result = self.block_builder.emit(
             relax.op.nn.adaptive_avg_pool2d(x, output_size, layout="NCHW")
         )
+        # Remove added batch dim from result
+        if x_ndim == 3:
+            result = relax.op.squeeze(result, axis=[0])
+        return result
+
+    def _adaptive_avg_pool3d(self, node: fx.Node) -> relax.Var:
+        x = self.env[node.args[0]]
+        output_size = node.args[1]
+        # Expand to 5D by adding batch dim if input is 4D
+        x_ndim = x.struct_info.ndim
+        if x_ndim == 4:
+            x = relax.op.expand_dims(x, axis=0)
+
+        result = self.block_builder.emit(
+            relax.op.nn.adaptive_avg_pool3d(x, output_size, layout="NCDHW")
+        )
+        # Remove added batch dim from result
+        if x_ndim == 4:
+            result = relax.op.squeeze(result, axis=[0])
+        return result
 
     def _addmm(self, node: fx.Node) -> relax.Var:
         x = self.env[node.args[0]]

python/tvm/relax/frontend/torch/exported_program_translator.py

@@ -387,7 +387,9 @@ def create_convert_map(
             "_native_batch_norm_legit_functional.default": self._batch_norm_legit_functional,
             "_native_batch_norm_legit_no_training.default": self._batch_norm_legit_no_training,
             "batch_norm.default": self._batch_norm_legit_no_training,
+            "adaptive_avg_pool1d.default": self._adaptive_avg_pool1d,
             "adaptive_avg_pool2d.default": self._adaptive_avg_pool2d,
+            "adaptive_avg_pool3d.default": self._adaptive_avg_pool3d,
             "addmm.default": self._addmm,
             "avg_pool2d.default": self._avg_pool2d,
             "baddbmm.default": self._baddbmm,

python/tvm/relax/frontend/torch/fx_translator.py

@@ -182,13 +182,53 @@ def call_binary_op(op, lhs, rhs):
 
     ########## Neural Network ##########
 
+    def _adaptive_avg_pool1d_module(self, node: fx.Node) -> relax.Var:
+        module = self.named_modules[node.target]
+        x = self.env[node.args[0]]
+        output_size = module.output_size
+        # Expand to 3D by adding batch dim if input is 2D
+        x_ndim = x.struct_info.ndim
+        if x_ndim == 2:
+            x = relax.op.expand_dims(x, axis=0)
+        result = self.block_builder.emit(
+            relax.op.nn.adaptive_avg_pool1d(x, output_size, layout="NCW")  # (N, C, L)
+        )
+        # Remove added batch dim from result
+        if x_ndim == 2:
+            result = relax.op.squeeze(result, axis=[0])
+        return result
+
     def _adaptive_avg_pool2d_module(self, node: fx.Node) -> relax.Var:
         module = self.named_modules[node.target]
         x = self.env[node.args[0]]
         output_size = module.output_size
-        return self.block_builder.emit(
+        # Expand to 4D by adding batch dim if input is 3D
+        x_ndim = x.struct_info.ndim
+        if x_ndim == 3:
+            x = relax.op.expand_dims(x, axis=0)
+        result = self.block_builder.emit(
             relax.op.nn.adaptive_avg_pool2d(x, output_size, layout="NCHW")
         )
+        # Remove added batch dim from result
+        if x_ndim == 3:
+            result = relax.op.squeeze(result, axis=[0])
+        return result
+
+    def _adaptive_avg_pool3d_module(self, node: fx.Node) -> relax.Var:
+        module = self.named_modules[node.target]
+        x = self.env[node.args[0]]
+        output_size = module.output_size
+        # Expand to 5D by adding batch dim if input is 4D
+        x_ndim = x.struct_info.ndim
+        if x_ndim == 4:
+            x = relax.op.expand_dims(x, axis=0)
+        result = self.block_builder.emit(
+            relax.op.nn.adaptive_avg_pool3d(x, output_size, layout="NCDHW")  # (N, C, D, H, W)
+        )
+        # Remove added batch dim from result
+        if x_ndim == 4:
+            result = relax.op.squeeze(result, axis=[0])
+        return result
 
     def _avg_pool2d_module(self, node: fx.Node) -> relax.Var:
         x = self.env[node.args[0]]
@@ -649,7 +689,9 @@ def create_convert_map(
             nn.Softplus: self._softplus_module,
             nn.Tanh: self._unary_op(relax.op.tanh),
             # neural network
+            nn.AdaptiveAvgPool1d: self._adaptive_avg_pool1d_module,
             nn.AdaptiveAvgPool2d: self._adaptive_avg_pool2d_module,
+            nn.AdaptiveAvgPool3d: self._adaptive_avg_pool3d_module,
             nn.AvgPool2d: self._avg_pool2d_module,
             nn.BatchNorm2d: self._batch_norm_2d_module,
             nn.Conv1d: self._conv1d_module,
@@ -755,7 +797,9 @@ def create_convert_map(
             "truediv": self._binary_op(relax.op.divide, operator.truediv),
             "xor": self._binary_op(relax.op.bitwise_xor, operator.xor),
             # neural network
+            "adaptive_avg_pool1d": self._adaptive_avg_pool1d,
             "adaptive_avg_pool2d": self._adaptive_avg_pool2d,
+            "adaptive_avg_pool3d": self._adaptive_avg_pool3d,
             "addmm": self._addmm,
             "avg_pool2d": self._avg_pool2d,
             "baddbmm": self._baddbmm,

tests/python/relax/test_frontend_from_exported_program.py

@@ -1145,6 +1145,38 @@ def main(
     verify_model(model, example_args, binding, expected1)
 
 
+def test_adaptive_avgpool1d():
+    class AdaptiveAvgPool1d0(torch.nn.Module):
+        def __init__(self):
+            super().__init__()
+            self.pool = torch.nn.AdaptiveAvgPool1d(output_size=5)
+
+        def forward(self, input):
+            return self.pool(input)
+
+    class AdaptiveAvgPool1d1(torch.nn.Module):
+        def forward(self, input):
+            return torch.nn.functional.adaptive_avg_pool1d(input, output_size=5)
+
+    @tvm.script.ir_module
+    class expected1:
+        @R.function
+        def main(
+            input_1: R.Tensor((1, 3, 10), dtype="float32")
+        ) -> R.Tuple(R.Tensor((1, 3, 5), dtype="float32")):
+            with R.dataflow():
+                lv: R.Tensor((1, 3, 5), dtype="float32") = R.nn.adaptive_avg_pool1d(
+                    input_1, output_size=[5], layout="NCW"
+                )
+                gv: R.Tuple(R.Tensor((1, 3, 5), dtype="float32")) = (lv,)
+                R.output(gv)
+            return gv
+
+    example_args = (torch.randn(1, 3, 10, dtype=torch.float32),)
+    verify_model(AdaptiveAvgPool1d0(), example_args, {}, expected1)
+    verify_model(AdaptiveAvgPool1d1(), example_args, {}, expected1)
+
+
 def test_adaptive_avgpool2d():
     class AdaptiveAvgPool2d0(Module):
         def __init__(self):
@@ -1178,6 +1210,38 @@ def main(
     verify_model(AdaptiveAvgPool2d1(), example_args, {}, expected1)
 
 
+def test_adaptive_avgpool3d():
+    class AdaptiveAvgPool3d0(torch.nn.Module):
+        def __init__(self):
+            super().__init__()
+            self.pool = torch.nn.AdaptiveAvgPool3d([4, 4, 4])
+
+        def forward(self, input):
+            return self.pool(input)
+
+    class AdaptiveAvgPool3d1(torch.nn.Module):
+        def forward(self, input):
+            return torch.nn.functional.adaptive_avg_pool3d(input, [4, 4, 4])
+
+    @tvm.script.ir_module
+    class expected1:
+        @R.function
+        def main(
+            input_1: R.Tensor((1, 3, 8, 8, 8), dtype="float32")
+        ) -> R.Tuple(R.Tensor((1, 3, 4, 4, 4), dtype="float32")):
+            with R.dataflow():
+                lv: R.Tensor((1, 3, 4, 4, 4), dtype="float32") = R.nn.adaptive_avg_pool3d(
+                    input_1, output_size=[4, 4, 4], layout="NCDHW", out_layout="NCDHW"
+                )
+                gv: R.Tuple(R.Tensor((1, 3, 4, 4, 4), dtype="float32")) = (lv,)
+                R.output(gv)
+            return gv
+
+    example_args = (torch.randn(1, 3, 8, 8, 8, dtype=torch.float32),)
+    verify_model(AdaptiveAvgPool3d0(), example_args, {}, expected1)
+    verify_model(AdaptiveAvgPool3d1(), example_args, {}, expected1)
+
+
 def test_addmm():
     class Addmm1(Module):
         def __init__(self):

tests/python/relax/test_frontend_from_fx.py

@@ -1181,6 +1181,39 @@ def main(input_1: R.Tensor((1, 3, 10, 10), dtype="float32")):
     verify_model(AvgPool2d4(), input_info, {}, expected3)
 
 
+def test_adaptive_avgpool1d():
+    input_info = [([1, 3, 16], "float32")]
+
+    class AdaptiveAvgPool1d0(torch.nn.Module):
+        def __init__(self):
+            super().__init__()
+            self.pool = torch.nn.AdaptiveAvgPool1d(8)
+
+        def forward(self, input):
+            return self.pool(input)
+
+    class AdaptiveAvgPool1d1(torch.nn.Module):
+        def forward(self, input):
+            return torch.nn.functional.adaptive_avg_pool1d(input, 8)
+
+    @tvm.script.ir_module
+    class expected1:
+        @R.function
+        def main(
+            input_1: R.Tensor((1, 3, 16), dtype="float32")
+        ) -> R.Tensor((1, 3, 8), dtype="float32"):
+            with R.dataflow():
+                lv: R.Tensor((1, 3, 8), dtype="float32") = R.nn.adaptive_avg_pool1d(
+                    input_1, output_size=[8], layout="NCW", out_layout="NCW"
+                )
+                gv: R.Tensor((1, 3, 8), dtype="float32") = lv
+                R.output(gv)
+            return gv
+
+    verify_model(AdaptiveAvgPool1d0(), input_info, {}, expected1)
+    verify_model(AdaptiveAvgPool1d1(), input_info, {}, expected1)
+
+
 def test_adaptive_avgpool2d():
     input_info = [([1, 3, 10, 10], "float32")]
 
@@ -1215,6 +1248,39 @@ def main(
     verify_model(AdaptiveAvgPool2d1(), input_info, {}, expected1)
 
 
+def test_adaptive_avgpool3d():
+    input_info = [([1, 3, 16, 16, 16], "float32")]
+
+    class AdaptiveAvgPool3d0(torch.nn.Module):
+        def __init__(self):
+            super().__init__()
+            self.pool = torch.nn.AdaptiveAvgPool3d((8, 8, 8))
+
+        def forward(self, input):
+            return self.pool(input)
+
+    class AdaptiveAvgPool3d1(torch.nn.Module):
+        def forward(self, input):
+            return torch.nn.functional.adaptive_avg_pool3d(input, (8, 8, 8))
+
+    @tvm.script.ir_module
+    class expected1:
+        @R.function
+        def main(
+            input_1: R.Tensor((1, 3, 16, 16, 16), dtype="float32")
+        ) -> R.Tensor((1, 3, 8, 8, 8), dtype="float32"):
+            with R.dataflow():
+                lv: R.Tensor((1, 3, 8, 8, 8), dtype="float32") = R.nn.adaptive_avg_pool3d(
+                    input_1, output_size=[8, 8, 8], layout="NCDHW", out_layout="NCDHW"
+                )
+                gv: R.Tensor((1, 3, 8, 8, 8), dtype="float32") = lv
+                R.output(gv)
+            return gv
+
+    verify_model(AdaptiveAvgPool3d0(), input_info, {}, expected1)
+    verify_model(AdaptiveAvgPool3d1(), input_info, {}, expected1)
+
+
 def test_flatten():
     input_info = [([1, 3, 10, 10], "float32")]