Support count_include_pad attr in avg_pool2d ONNX (tracel-ai#978)

Fixes tracel-ai#636
theparitt · Nov 21, 2023 · 445f41b · 445f41b
1 parent cb616ed
commit 445f41b
Show file tree

Hide file tree

Showing 6 changed files with 58 additions and 22 deletions.
diff --git a/burn-core/src/nn/pool/avg_pool2d.rs b/burn-core/src/nn/pool/avg_pool2d.rs
@@ -20,7 +20,7 @@ pub struct AvgPool2dConfig {
     pub padding: PaddingConfig2d,
     /// If the padding is counted in the denominator when computing the average.
     #[config(default = "true")]
-    count_include_pad: bool,
+    pub count_include_pad: bool,
 }
 
 /// Applies a 2D avg pooling over input tensors.

diff --git a/burn-import/onnx-tests/tests/avg_pool2d/avg_pool2d.onnx b/burn-import/onnx-tests/tests/avg_pool2d/avg_pool2d.onnx
diff --git a/burn-import/onnx-tests/tests/avg_pool2d/avg_pool2d.py b/burn-import/onnx-tests/tests/avg_pool2d/avg_pool2d.py
@@ -10,14 +10,20 @@ class Model(nn.Module):
     def __init__(self):
         super(Model, self).__init__()
 
-        # TODO when https://github.com/burn-rs/burn/issues/636 is resolved, test this with a model 
-        # that uses `count_include_pad=False` and padding=(2, 1) 
-        self.pool2d = nn.AvgPool2d((4, 2), stride=(
-            2, 1), padding=(0, 0), count_include_pad=False)
+        self.pool2d1 = nn.AvgPool2d((4, 2), stride=(
+            2, 1))
 
-    def forward(self, x):
-        x = self.pool2d(x)
-        return x
+        self.pool2d2 = nn.AvgPool2d((4, 2), stride=(
+            2, 1), padding=(2, 1), count_include_pad=True)
+
+        self.pool2d3 = nn.AvgPool2d((4, 2), stride=(
+            2, 1), padding=(2, 1), count_include_pad=False)
+
+    def forward(self, x1, x2, x3):
+        y1 = self.pool2d1(x1)
+        y2 = self.pool2d2(x2)
+        y3 = self.pool2d3(x3)
+        return y1, y2, y3
 
 
 def main():
@@ -33,18 +39,22 @@ def main():
     device = torch.device("cpu")
 
     file_name = "avg_pool2d.onnx"
-    test_input = torch.randn(1, 1, 5, 5, device=device)
-    torch.onnx.export(model, test_input, file_name,
+    input1 = torch.randn(1, 1, 5, 5, device=device)
+    torch.onnx.export(model, (input1, input1, input1), file_name,
                       verbose=False, opset_version=16)
 
     print("Finished exporting model to {}".format(file_name))
 
     # Output some test data for use in the test
-    print("Test input data shape of ones: {}".format(test_input.shape))
-    print("Test input data of ones: {}".format(test_input))
-    output = model.forward(test_input)
-    print("Test output data shape: {}".format(output.shape))
-    print("Test output: {}".format(output))
+    print("Test input data shape: {}".format(input1.shape))
+    print("Test input data: {}".format(input1))
+    output1, output2, output3 = model.forward(input1, input1, input1)
+    print("Test output1 data shape: {}".format(output1.shape))
+    print("Test output2 data shape: {}".format(output2.shape))
+    print("Test output3 data shape: {}".format(output3.shape))
+    print("Test output1: {}".format(output1))
+    print("Test output2: {}".format(output2))
+    print("Test output3: {}".format(output3))
 
 
 if __name__ == '__main__':

diff --git a/burn-import/onnx-tests/tests/onnx_tests.rs b/burn-import/onnx-tests/tests/onnx_tests.rs
@@ -362,10 +362,30 @@ mod tests {
             [-1.805, -0.476, 0.205, 0.338, 1.353],
             [0.374, 0.013, 0.774, -0.109, -0.271],
         ]]]);
-        let output = model.forward(input);
-        let expected = Data::from([[[[0.008, -0.131, -0.208, 0.425]]]]);
+        let (output1, output2, output3) = model.forward(input.clone(), input.clone(), input);
+        let expected1 = Data::from([[[[0.008, -0.131, -0.208, 0.425]]]]);
+        let expected2 = Data::from([[[
+            [-0.045, 0.202, -0.050, -0.295, 0.162, 0.160],
+            [-0.176, 0.008, -0.131, -0.208, 0.425, 0.319],
+            [-0.084, -0.146, 0.017, 0.170, 0.216, 0.125],
+        ]]]);
+        let expected3 = Data::from([[[
+            [-0.182, 0.404, -0.100, -0.590, 0.324, 0.638],
+            [-0.352, 0.008, -0.131, -0.208, 0.425, 0.638],
+            [-0.224, -0.195, 0.023, 0.226, 0.288, 0.335],
+        ]]]);
+
+        let expected_shape1 = Shape::from([1, 1, 1, 4]);
+        let expected_shape2 = Shape::from([1, 1, 3, 6]);
+        let expected_shape3 = Shape::from([1, 1, 3, 6]);
+
+        assert_eq!(output1.shape(), expected_shape1);
+        assert_eq!(output2.shape(), expected_shape2);
+        assert_eq!(output3.shape(), expected_shape3);
 
-        output.to_data().assert_approx_eq(&expected, 3);
+        output1.to_data().assert_approx_eq(&expected1, 3);
+        output2.to_data().assert_approx_eq(&expected2, 3);
+        output3.to_data().assert_approx_eq(&expected3, 3);
     }
 
     #[test]

diff --git a/burn-import/src/burn/node/avg_pool2d.rs b/burn-import/src/burn/node/avg_pool2d.rs
@@ -51,6 +51,7 @@ impl<PS: PrecisionSettings> NodeCodegen<PS> for AvgPool2dNode {
         let kernel_size = self.config.kernel_size.to_tokens();
         let strides = self.config.strides.to_tokens();
         let padding = self.config.padding.to_tokens();
+        let count_include_pad = self.config.count_include_pad;
 
         let init_line = quote! {
             init();
@@ -60,6 +61,7 @@ impl<PS: PrecisionSettings> NodeCodegen<PS> for AvgPool2dNode {
             let #name = AvgPool2dConfig::new(#kernel_size)
                 .with_strides(#strides)
                 .with_padding(#padding)
+                .with_count_include_pad(#count_include_pad)
                 .#init_line
         };
 
@@ -137,6 +139,7 @@ mod tests {
                     let avg_pool2d = AvgPool2dConfig::new([3, 3])
                         .with_strides([1, 1])
                         .with_padding(PaddingConfig2d::Valid)
+                        .with_count_include_pad(true)
                         .init();
 
                     Self {

diff --git a/burn-import/src/onnx/op_configuration.rs b/burn-import/src/onnx/op_configuration.rs
@@ -129,26 +129,29 @@ pub fn avg_pool2d_config(curr: &Node) -> AvgPool2dConfig {
     let mut strides = vec![1, 1];
     let mut pads = vec![0, 0, 0, 0];
     let mut count_include_pad: i64 = 0;
+    let mut ceil_mode: i64 = 0;
 
     for (key, value) in curr.attrs.iter() {
         match key.as_str() {
             "kernel_shape" => kernel_shape = value.clone().into_i64s(),
             "strides" => strides = value.clone().into_i64s(),
             "pads" => pads = value.clone().into_i64s(),
             "count_include_pad" => count_include_pad = value.clone().into_i64(),
+            "ceil_mode" => ceil_mode = value.clone().into_i64(),
             _ => {}
         }
     }
 
-    let padding = padding_config(&pads);
-
-    if count_include_pad == 1 && padding != PaddingConfig2d::Valid {
-        todo!("AvgPool2d: count_include_pad is not supported. See https://github.com/burn-rs/burn/issues/636");
+    if ceil_mode == 1 {
+        panic!("ceil_mode is not supported");
     }
 
+    let padding = padding_config(&pads);
+
     AvgPool2dConfig::new([kernel_shape[0] as usize, kernel_shape[1] as usize])
         .with_strides([strides[0] as usize, strides[1] as usize])
         .with_padding(padding)
+        .with_count_include_pad(count_include_pad == 1)
 }
 
 /// Create a FlattenConfig from the attributes of the node