提交 20571503 编写于 作者: C Channingss

fix bug

上级 bcf4b473
......@@ -616,7 +616,7 @@ class PaddleOpMapper(object):
outputs=[cast_shape_name],
to=onnx_pb.TensorProto.INT64)
node_list.extend([concat_shape_node, cast_shape_node])
shape_name3 = op.output('Out')[0] + "@shape.concat44"
shape_name3 = op.output('Out')[0] + "@shape.concat3"
shape_node3 = helper.make_node(
'Concat',
inputs=[shape_name1, cast_shape_name],
......@@ -641,9 +641,16 @@ class PaddleOpMapper(object):
inputs=[shape_name3, name_h_w],
outputs=[outputs_h_w_scales])
node_list.append(node_h_w_scales)
shape_name4 = op.output('Out')[0] + "@shape.concat4"
shape_node4 = helper.make_node(
'Concat',
inputs=[shape_name1, outputs_h_w_scales],
outputs=[shape_name4],
axis=0)
node_list.append(shape_node4)
result_node = helper.make_node(
'Resize',
inputs=[op.input('X')[0], outputs_h_w_scales],
inputs=[op.input('X')[0], shape_name4],
outputs=op.output('Out'),
mode='linear', )
node_list.extend([result_node])
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册