Dyn gather (#1513)

Dynamic shape support for gather op.

src/include/migraphx/op/gather.hpp

@@ -26,6 +26,7 @@
 
 #include <array>
 #include <migraphx/check_shapes.hpp>
+#include <migraphx/dyn_output.hpp>
 #include <migraphx/stringutils.hpp>
 #include <migraphx/streamutils.hpp>
 #include <migraphx/literal.hpp>
@@ -61,35 +62,59 @@ struct gather
 
     shape normalize_compute_shape(std::vector<shape> inputs) const
     {
-        check_shapes{inputs, *this}.has(2);
-        auto lens = inputs[0].lens();
-        auto type = inputs[0].type();
-        lens.erase(lens.begin() + axis);
-        if(not inputs[1].scalar())
+        check_shapes{inputs, *this, true}.has(2);
+        shape data    = inputs[0];
+        shape indices = inputs[1];
+        auto type     = data.type();
+        // If index_dims is dynamic, convert the data to dynamic too.
+        if(indices.dynamic())
         {
-            auto ind_lens = inputs[1].lens();
-            lens.insert(lens.begin() + axis, ind_lens.begin(), ind_lens.end());
+            data = data.to_dynamic();
         }
-
-        // for scalar output
-        if(lens.empty())
+        if(data.dynamic())
         {
-            return {type};
+            auto dims = data.dyn_dims();
+            dims.erase(dims.begin() + axis);
+
+            if(not indices.scalar())
+            {
+                auto index_dims = indices.to_dynamic().dyn_dims();
+                dims.insert(dims.begin() + axis, index_dims.begin(), index_dims.end());
+            }
+            return {type, dims};
         }
+        else
+        {
+            // Both data and indices are static.  indices may be scalar
+            auto lens = data.lens();
+            lens.erase(lens.begin() + axis);
 
-        return {type, lens};
+            if(not indices.scalar())
+            {
+                auto ind_lens = indices.lens();
+                lens.insert(lens.begin() + axis, ind_lens.begin(), ind_lens.end());
+            }
+
+            // for scalar output
+            if(lens.empty())
+            {
+                return {type};
+            }
+
+            return {type, lens};
+        }
     }
 
-    argument compute(const shape& output_shape, std::vector<argument> args) const
+    argument compute(const dyn_output& dyn_out, std::vector<argument> args) const
     {
-        argument result{output_shape};
+        argument result{dyn_out.computed_shape};
         // negative axis means counting dimensions from back
         auto lens                 = args[0].get_shape().lens();
         std::size_t axis_dim_size = lens[axis];
         // max dimension in axis
         visit_all(result, args[0])([&](auto output, auto data) {
             args[1].visit([&](auto indices) {
-                if(output_shape.scalar())
+                if(dyn_out.computed_shape.scalar())
                 {
                     auto in_index = indices.front();
                     in_index      = (in_index < 0) ? in_index + axis_dim_size : in_index;

test/onnx/gen_onnx.py

@@ -2053,6 +2053,40 @@ def gather_test():
     return ([node], [x, i], [y])
 
 
+@onnx_test()
+def gather_scalar_test():
+    x = helper.make_tensor_value_info('data', TensorProto.FLOAT, [3, 4, 5, 6])
+    i = helper.make_tensor_value_info('indices', TensorProto.INT32, [])
+    y = helper.make_tensor_value_info('y', TensorProto.FLOAT, [4, 5, 6])
+
+    node = onnx.helper.make_node(
+        'Gather',
+        inputs=['data', 'indices'],
+        outputs=['y'],
+        axis=1,
+    )
+
+    return ([node], [x, i], [y])
+
+
+@onnx_test()
+def gather_dyn_test():
+    x = helper.make_tensor_value_info('data', TensorProto.FLOAT,
+                                      [None, 4, 5, 6])
+    i = helper.make_tensor_value_info('indices', TensorProto.INT32,
+                                      [None, 3, 4, 5])
+    y = helper.make_tensor_value_info('y', TensorProto.FLOAT, [2, 3, 4, 5])
+
+    node = onnx.helper.make_node(
+        'Gather',
+        inputs=['data', 'indices'],
+        outputs=['y'],
+        axis=1,
+    )
+
+    return ([node], [x, i], [y])
+
+
 @onnx_test()
 def gather_elements_axis0_test():
     x = helper.make_tensor_value_info('data', TensorProto.FLOAT, [3, 4])

test/onnx/onnx_test.cpp

@@ -2048,6 +2048,46 @@ TEST_CASE(gather_test)
     EXPECT(p == prog);
 }
 
+TEST_CASE(gather_scalar_test)
+{
+    migraphx::program p;
+    auto* mm = p.get_main_module();
+    auto l0 = mm->add_parameter("data", migraphx::shape{migraphx::shape::float_type, {3, 4, 5, 6}});
+    std::vector<size_t> idims{1};
+    auto l1 =
+        mm->add_parameter("indices", migraphx::shape{migraphx::shape::int32_type, idims, {0}});
+    int axis = 1;
+    mm->add_instruction(migraphx::make_op("gather", {{"axis", axis}}), l0, l1);
+    auto prog = optimize_onnx("gather_scalar_test.onnx");
+
+    EXPECT(p == prog);
+}
+
+TEST_CASE(gather_dyn_test)
+{
+    migraphx::program p;
+    auto* mm = p.get_main_module();
+    auto l0  = mm->add_parameter(
+        "data",
+        migraphx::shape{migraphx::shape::float_type, {{1, 4, 0}, {4, 4, 0}, {5, 5, 0}, {6, 6, 0}}});
+    auto l1 = mm->add_parameter(
+        "indices",
+        migraphx::shape{migraphx::shape::int32_type, {{1, 4, 0}, {3, 3, 0}, {4, 4, 0}, {5, 5, 0}}});
+    auto cont_l0 = mm->add_instruction(migraphx::make_op("contiguous"), l0);
+    auto cont_l1 = mm->add_instruction(migraphx::make_op("contiguous"), l1);
+
+    int axis       = 1;
+    auto gather_op = migraphx::make_op("gather", {{"axis", axis}});
+    auto ret       = mm->add_instruction(gather_op, cont_l0, cont_l1);
+    mm->add_return({ret});
+
+    migraphx::onnx_options options;
+    options.default_dyn_dim_value = {1, 4, 0};
+    auto prog                     = parse_onnx("gather_dyn_test.onnx", options);
+
+    EXPECT(p == prog);
+}
+
 TEST_CASE(gather_elements_axis0_test)
 {
     migraphx::program p;

test/op_shape_test.cpp

@@ -831,6 +831,77 @@ TEST_CASE(gather)
     }
 }
 
+TEST_CASE(gather_dyn0)
+{
+    // Insert dynamic index into dynamic shape
+    migraphx::shape input{migraphx::shape::float_type,
+                          {{2, 3, 2}, {3, 4, 3}, {6, 9, 7}, {12, 14, 13}}};
+    migraphx::shape indices{migraphx::shape::int32_type, {{2, 7, 3}, {3, 3, 0}}};
+    int axis = 1;
+    expect_shape(migraphx::shape{migraphx::shape::float_type,
+                                 {{2, 3, 2}, {2, 7, 3}, {3, 3, 0}, {6, 9, 7}, {12, 14, 13}}},
+                 migraphx::make_op("gather", {{"axis", axis}}),
+                 input,
+                 indices);
+}
+
+TEST_CASE(gather_dyn1)
+{
+    // Insert static index into dynamic shape
+    migraphx::shape input{migraphx::shape::float_type,
+                          {{2, 3, 2}, {3, 4, 3}, {6, 9, 7}, {12, 14, 13}}};
+    migraphx::shape indices{migraphx::shape::int32_type, {2, 3}};
+    int axis = 1;
+    expect_shape(migraphx::shape{migraphx::shape::float_type,
+                                 {{2, 3, 2}, {2, 2, 0}, {3, 3, 0}, {6, 9, 7}, {12, 14, 13}}},
+                 migraphx::make_op("gather", {{"axis", axis}}),
+                 input,
+                 indices);
+}
+
+TEST_CASE(gather_dyn2)
+{
+    // Insert scalar (static) index into dynamic shape
+    migraphx::shape input{migraphx::shape::float_type,
+                          {{2, 3, 2}, {3, 4, 3}, {6, 9, 7}, {12, 14, 13}}};
+
+    std::vector<std::size_t> mins;
+    std::vector<std::size_t> maxes;
+    std::vector<std::size_t> opts;
+    migraphx::shape indices{migraphx::shape::int32_type, mins, maxes, opts};
+    int axis = 1;
+    expect_shape(migraphx::shape{migraphx::shape::float_type, {{2, 3, 2}, {6, 9, 7}, {12, 14, 13}}},
+                 migraphx::make_op("gather", {{"axis", axis}}),
+                 input,
+                 indices);
+}
+
+TEST_CASE(gather_dyn3)
+{
+    // Insert dynamic index into static shape, axis 1
+    migraphx::shape input{migraphx::shape::float_type, {2, 3, 6, 12}};
+    migraphx::shape indices{migraphx::shape::int32_type, {{2, 3, 2}, {3, 4, 3}}};
+    int axis = 1;
+    expect_shape(migraphx::shape{migraphx::shape::float_type,
+                                 {{2, 2, 0}, {2, 3, 2}, {3, 4, 3}, {6, 6, 0}, {12, 12, 0}}},
+                 migraphx::make_op("gather", {{"axis", axis}}),
+                 input,
+                 indices);
+}
+
+TEST_CASE(gather_dyn4)
+{
+    // Insert dynamic index into static shape, axis 0
+    migraphx::shape input{migraphx::shape::float_type, {2, 3, 6, 12}};
+    migraphx::shape indices{migraphx::shape::int32_type, {{2, 3, 2}, {3, 4, 3}}};
+    int axis = 0;
+    expect_shape(migraphx::shape{migraphx::shape::float_type,
+                                 {{2, 3, 2}, {3, 4, 3}, {3, 3, 0}, {6, 6, 0}, {12, 12, 0}}},
+                 migraphx::make_op("gather", {{"axis", axis}}),
+                 input,
+                 indices);
+}
+
 TEST_CASE(get_tuple_elem_test)
 {
     migraphx::shape s0{migraphx::shape::bool_type, {1, 1}};

test/ref_ops_test.cpp

@@ -2524,6 +2524,78 @@ TEST_CASE(gather_test)
     }
 }
 
+TEST_CASE(gather_dyn_test0)
+{
+    // Dynamic data, static indices
+    migraphx::program p;
+    auto* mm = p.get_main_module();
+    migraphx::shape s{migraphx::shape::int32_type, {{2, 5, 0}, {3, 3, 0}}};
+
+    auto x = mm->add_parameter("x", s);
+    std::vector<int> indices{1, 2};
+
+    migraphx::shape s_ind{migraphx::shape::int32_type, {1, 2}};
+    auto ind = mm->add_parameter("indices", s_ind);
+    mm->add_instruction(migraphx::make_op("gather", {{"axis", 1}}), x, ind);
+
+    migraphx::shape sresult{migraphx::shape::int32_type, {{2, 5, 0}, {1, 1, 0}, {2, 2, 0}}};
+    EXPECT(p.get_output_shapes().back() == sresult);
+    p.compile(migraphx::ref::target{});
+
+    migraphx::shape input_fixed_shape{migraphx::shape::int32_type, {2, 3}};
+    migraphx::shape input_indices{migraphx::shape::int32_type, {1, 2}};
+    migraphx::parameter_map params;
+    std::vector<int> data(2 * 3);
+    std::iota(data.begin(), data.end(), 0);
+    params["x"]       = migraphx::argument(input_fixed_shape, data.data());
+    params["indices"] = migraphx::argument(input_indices, indices.data());
+    auto result       = p.eval(params).back();
+
+    std::vector<int> gold = {1, 2, 4, 5};
+    std::vector<int> results_vector(2 * 1 * 2);
+    result.visit([&](auto output) { results_vector.assign(output.begin(), output.end()); });
+    EXPECT(migraphx::verify_range(results_vector, gold));
+    migraphx::shape sfinal{migraphx::shape::int32_type, {2, 1, 2}};
+    EXPECT(result.get_shape() == sfinal);
+}
+
+TEST_CASE(gather_dyn_test1)
+{
+    // Dynamic data, dynamic indices
+    migraphx::program p;
+    auto* mm = p.get_main_module();
+    migraphx::shape s{migraphx::shape::int32_type, {{2, 5, 0}, {4, 4, 0}}};
+
+    auto x = mm->add_parameter("x", s);
+
+    migraphx::shape s_ind{migraphx::shape::int32_type, {{1, 8, 7}, {2, 3, 3}}};
+    auto ind = mm->add_parameter("indices", s_ind);
+    mm->add_instruction(migraphx::make_op("gather", {{"axis", 0}}), x, ind);
+
+    migraphx::shape sresult{migraphx::shape::int32_type, {{1, 8, 7}, {2, 3, 3}, {4, 4, 0}}};
+    EXPECT(p.get_output_shapes().back() == sresult);
+    p.compile(migraphx::ref::target{});
+
+    migraphx::shape input_fixed_shape{migraphx::shape::int32_type, {3, 4}};
+    migraphx::shape input_indices_shape{migraphx::shape::int32_type, {1, 2}};
+    std::vector<int> indices{2, 0};
+    migraphx::parameter_map params;
+
+    std::vector<int> data(3 * 4);
+    std::iota(data.begin(), data.end(), 0);
+    params["x"]       = migraphx::argument(input_fixed_shape, data.data());
+    params["indices"] = migraphx::argument(input_indices_shape, indices.data());
+    auto result       = p.eval(params).back();
+
+    std::vector<int> gold = {8, 9, 10, 11, 0, 1, 2, 3};
+    std::vector<int> results_vector(1 * 2 * 4);
+    result.visit([&](auto output) { results_vector.assign(output.begin(), output.end()); });
+
+    EXPECT(migraphx::verify_range(results_vector, gold));
+    migraphx::shape sfinal{migraphx::shape::int32_type, {1, 2, 4}};
+    EXPECT(result.get_shape() == sfinal);
+}
+
 TEST_CASE(gathernd_test)
 {
     {