Add tile op (#513)

* fix pad calc

* add tile op

* formatting

* add test

* formatting

* fix tidy

* formatting
Co-authored-by: mvermeulen <5479696+mvermeulen@users.noreply.github.com>

src/onnx/onnx.cpp

@@ -131,6 +131,7 @@ struct onnx_parser
         add_mem_op("Softmax", &onnx_parser::parse_softmax<op::softmax>);
         add_mem_op("Split", &onnx_parser::parse_split);
         add_mem_op("Squeeze", &onnx_parser::parse_squeeze);
+        add_mem_op("Tile", &onnx_parser::parse_tile);
         add_mem_op("Transpose", &onnx_parser::parse_transpose);
         add_mem_op("Unsqueeze", &onnx_parser::parse_unsqueeze);
         add_mem_op("LSTM", &onnx_parser::parse_lstm);
@@ -1885,6 +1886,26 @@ struct onnx_parser
         return prog.add_instruction(op::add{}, l_mul, unsq_off_val);
     }
 
+    instruction_ref
+    parse_tile(const std::string&, const node_info&, std::vector<instruction_ref> args)
+    {
+        migraphx::argument arg_s = args[1]->eval();
+        check_arg_empty(arg_s, "PARSE_TILE: dynamic shape is not supported");
+        std::vector<std::int64_t> repeats;
+        arg_s.visit([&](auto input) { repeats.assign(input.begin(), input.end()); });
+
+        auto l0 = args[0];
+        for(int i = 0; i < repeats.size(); i++)
+        {
+            auto l1 = l0;
+            for(int j = 1; j < repeats[i]; j++)
+            {
+                l0 = prog.add_instruction(op::concat{i}, l0, l1);
+            }
+        }
+        return l0;
+    }
+
     void parse_from(std::istream& is)
     {
         onnx::ModelProto model;

test/onnx/gen_onnx.py

@@ -2150,6 +2150,30 @@ def tanh_test():
     return ([node], [x], [y])
 
 
+@onnx_test
+def tile_test():
+    x = helper.make_tensor_value_info('x', TensorProto.FLOAT, [2, 2])
+    y = helper.make_tensor_value_info('y', TensorProto.INT64, [2])
+    z = helper.make_tensor_value_info('z', TensorProto.FLOAT, [2, 4])
+
+    node = onnx.helper.make_node('Tile', inputs=['x', 'y'], outputs=['z'])
+
+    return ([node], [x, y], [z],
+            [helper.make_tensor('y', TensorProto.INT64, [2], [1, 2])])
+
+
+@onnx_test
+def tile_test_3x2():
+    x = helper.make_tensor_value_info('x', TensorProto.FLOAT, [2, 2])
+    y = helper.make_tensor_value_info('y', TensorProto.INT64, [2])
+    z = helper.make_tensor_value_info('z', TensorProto.FLOAT, [6, 4])
+
+    node = onnx.helper.make_node('Tile', inputs=['x', 'y'], outputs=['z'])
+
+    return ([node], [x, y], [z],
+            [helper.make_tensor('y', TensorProto.INT64, [2], [3, 2])])
+
+
 @onnx_test
 def transpose_test():
     x = helper.make_tensor_value_info('0', TensorProto.FLOAT, [1, 2, 2, 3])

test/onnx/onnx_test.cpp

@@ -1628,6 +1628,32 @@ TEST_CASE(tanh_test)
     EXPECT(p == prog);
 }
 
+TEST_CASE(tile_test)
+{
+    migraphx::program p;
+    p.add_literal(migraphx::literal{migraphx::shape{migraphx::shape::int64_type, {2}}, {1, 2}});
+    auto input = p.add_parameter("x", migraphx::shape{migraphx::shape::float_type, {2, 2}});
+    p.add_instruction(migraphx::op::concat{1}, input, input);
+
+    auto prog = optimize_onnx("tile_test.onnx");
+
+    EXPECT(p == prog);
+}
+
+TEST_CASE(tile_test_3x2)
+{
+    migraphx::program p;
+    p.add_literal(migraphx::literal{migraphx::shape{migraphx::shape::int64_type, {2}}, {3, 2}});
+    auto input = p.add_parameter("x", migraphx::shape{migraphx::shape::float_type, {2, 2}});
+    auto l0    = p.add_instruction(migraphx::op::concat{0}, input, input);
+    auto l1    = p.add_instruction(migraphx::op::concat{0}, l0, input);
+    p.add_instruction(migraphx::op::concat{1}, l1, l1);
+
+    auto prog = optimize_onnx("tile_test_3x2.onnx");
+
+    EXPECT(p == prog);
+}
+
 TEST_CASE(transpose_test)
 {
     migraphx::program p;

test/onnx/tile_test.onnx

+	tile_test:d
+
+
x
+
y
z"Tile	tile_test*:
B
yZ
+
x
+

+
+Z
+
y
+
+
+b
+
z
+

+
+B

test/onnx/tile_test_3x2.onnx

+
tile_test_3x2:h
+
+
x
+
y
z"Tile
tile_test_3x2*:B
yZ
+
x
+

+
+Z
+
y
+
+
+b
+
z
+

+
+B