Explicit padding for const shape ConvTranspose (#1513)

* explicit padding for const shape ConvTranspose Signed-off-by: Calvin McCarter <calvin@lightmatter.co> * pylint fix Signed-off-by: Calvin McCarter <calvin@lightmatter.co> * test_backend graph_validator fix Signed-off-by: Calvin McCarter <calvin@lightmatter.co> * keep output_shape if input_dims unknown Signed-off-by: Calvin McCarter <calvin@lightmatter.co> * cleanup Signed-off-by: Calvin McCarter <calvin@lightmatter.co> * remove unused variable Signed-off-by: Calvin McCarter <calvin@lightmatter.co> Co-authored-by: TomWildenhain-Microsoft <67606533+TomWildenhain-Microsoft@users.noreply.github.com> Co-authored-by: Guenther Schmuelling <guschmue@microsoft.com>
onnx · May 13, 2021 · 620b952 · 620b952
1 parent 46cb038
commit 620b952
Show file tree

Hide file tree

Showing 2 changed files with 30 additions and 5 deletions.
diff --git a/tests/test_backend.py b/tests/test_backend.py
@@ -4706,7 +4706,7 @@ def func(filter_val, out_backprop_val, batch_dim):
         def graph_validator(g):
             for n in g.get_nodes():
                 if n.type == 'ConvTranspose':
-                    return "output_shape" in n.attr
+                    return "pads" in n.attr or "output_shape" in n.attr
             return False
         self._run_test_case(func, [_OUTPUT], {_INPUT: filters_val, _INPUT1: out_backprop_val, _INPUT2: batch_dim_val},
                             graph_validator=graph_validator)

diff --git a/tf2onnx/onnx_opset/nn.py b/tf2onnx/onnx_opset/nn.py
@@ -437,10 +437,9 @@ def version_1(cls, ctx, node, **kwargs):
             input_dims = input_shape[2:2+spatial]
         output_shape_orig = node.output_shapes
 
-        # ouput_shape is explicitly specified here, in this case pads values are auto generated/calculated.
+        # output_shape is explicitly specified here and then converted to explicit pads.
         output_shape = get_shape_from_const_or_concat(ctx, node.inputs[0])
         if output_shape is not None:
-            #output_shape = ctx.get_shape(node.output[0])
             if is_channels_last(node):
                 new_output_shape = [output_shape[1], output_shape[2]]
                 if spatial == 3:
@@ -453,8 +452,34 @@ def version_1(cls, ctx, node, **kwargs):
             utils.make_sure(new_output_shape.count(-1) <= 0, "output dims need to be known")
             utils.make_sure(all(new_output_shape[i] >= input_dims[i] for i in range(spatial)),
                             "output dims cannot be smaller than input dims.")
-
-            node.set_attr("output_shape", new_output_shape)
+            if -1 in input_dims:
+                node.set_attr("output_shape", new_output_shape)
+            else:
+                if "strides" in node.attr:
+                    strides = parse_dims_attr(node, node.get_attr("strides").ints, spatial)
+                else:
+                    strides = [1] * spatial
+                if "dilations" in node.attr:
+                    dilations = parse_dims_attr(node, node.get_attr("dilations").ints, spatial)
+                else:
+                    dilations = [1] * spatial
+                if "output_padding" in node.attr:
+                    output_padding = parse_dims_attr(node, node.get_attr("output_padding").ints, spatial)
+                else:
+                    output_padding = [0] * spatial
+                kernel_shape = parse_dims_attr(node, node.get_attr("kernel_shape").ints, spatial)
+                total_padding = [-1] * spatial
+                pads = [1] * (spatial * 2)
+                for i in range(spatial):
+                    total_padding[i] = (strides[i] * (input_dims[i] - 1) + output_padding[i]
+                                        + ((kernel_shape[i] - 1) * dilations[i] + 1)
+                                        - new_output_shape[i])
+                    start_i = i
+                    end_i = i + spatial
+                    pads[start_i] = int(total_padding[i] / 2)
+                    pads[end_i] = total_padding[i] - pads[start_i]
+                node.set_attr("pads", pads)
+                node.set_attr("auto_pad", "NOTSET")
         else:
             utils.make_sure(ctx.opset >= 10, "Opset 10 needed for Conv Backprop Input with non-constant shape")
             strides = parse_dims_attr(node, node.get_attr('strides').ints, spatial)