Improve API for program parameters (#635)

* Take numpy array directly in python API

* Formatting

* Intialize program parameters from initializer list

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

src/api/include/migraphx/migraphx.hpp

@@ -379,6 +379,13 @@ struct program_parameters : MIGRAPHX_HANDLE_BASE(program_parameters)
 
     program_parameters() { this->make_handle(&migraphx_program_parameters_create); }
 
+    program_parameters(std::initializer_list<std::pair<std::string, argument>> l)
+    {
+        this->make_handle(&migraphx_program_parameters_create);
+        for(auto&& p : l)
+            this->add(p.first.c_str(), p.second);
+    }
+
     void add(const char* pname, const argument& pargument) const
     {
         call(&migraphx_program_parameters_add,

src/py/migraphx_py.cpp

@@ -180,7 +180,18 @@ PYBIND11_MODULE(migraphx, m)
              },
              py::arg("t"),
              py::arg("offload_copy") = true)
-        .def("run", &migraphx::program::eval)
+        .def("run",
+             [](migraphx::program& p, py::dict params) {
+                 migraphx::program::parameter_map pm;
+                 for(auto x : params)
+                 {
+                     std::string key      = x.first.cast<std::string>();
+                     py::buffer b         = x.second.cast<py::buffer>();
+                     py::buffer_info info = b.request();
+                     pm[key]              = migraphx::argument(to_shape(info), info.ptr);
+                 }
+                 return p.eval(pm);
+             })
         .def("sort", &migraphx::program::sort)
         .def("__eq__", std::equal_to<migraphx::program>{})
         .def("__ne__", std::not_equal_to<migraphx::program>{})

test/api/test_cpu.cpp

@@ -22,6 +22,25 @@ TEST_CASE(load_and_run)
     CHECK(bool{shapes_before.front() == outputs.front().get_shape()});
 }
 
+TEST_CASE(load_and_run_init_list)
+{
+    auto p             = migraphx::parse_onnx("conv_relu_maxpool_test.onnx");
+    auto shapes_before = p.get_output_shapes();
+    p.compile(migraphx::target("cpu"));
+    auto shapes_after = p.get_output_shapes();
+    CHECK(shapes_before.size() == 1);
+    CHECK(shapes_before.size() == shapes_after.size());
+    CHECK(bool{shapes_before.front() == shapes_after.front()});
+    auto param_shapes = p.get_parameter_shapes();
+    EXPECT(param_shapes.size() == 3);
+    auto names   = param_shapes.names();
+    auto outputs = p.eval({{names[0], migraphx::argument::generate(param_shapes[names[0]])},
+                           {names[1], migraphx::argument::generate(param_shapes[names[1]])},
+                           {names[2], migraphx::argument::generate(param_shapes[names[2]])}});
+    CHECK(shapes_before.size() == outputs.size());
+    CHECK(bool{shapes_before.front() == outputs.front().get_shape()});
+}
+
 TEST_CASE(quantize_fp16)
 {
     auto p1        = migraphx::parse_onnx("gemm_ex_test.onnx");

test/py/test_gpu.py

@@ -31,11 +31,9 @@ def test_sub_uint64():
     params = {}
 
     shapes = p.get_parameter_shapes()
-    arg0 = np.arange(120).reshape(shapes["0"].lens()).astype(np.uint64)
-    arg1 = np.arange(20).reshape(shapes["1"].lens()).astype(np.uint64)
+    params["0"] = np.arange(120).reshape(shapes["0"].lens()).astype(np.uint64)
+    params["1"] = np.arange(20).reshape(shapes["1"].lens()).astype(np.uint64)
 
-    params["0"] = migraphx.argument(arg0)
-    params["1"] = migraphx.argument(arg1)
     r = p.run(params)
     print(r)
 
@@ -49,9 +47,8 @@ def test_neg_int64():
     params = {}
 
     shapes = p.get_parameter_shapes()
-    arg0 = np.arange(6).reshape(shapes["0"].lens()).astype(np.int64)
+    params["0"] = np.arange(6).reshape(shapes["0"].lens()).astype(np.int64)
 
-    params["0"] = migraphx.argument(arg0)
     r = p.run(params)
     print(r)
 
@@ -68,8 +65,8 @@ def test_fp16_imagescaler():
 
     params = {}
     shapes = p.get_parameter_shapes()
-    arg0 = np.random.randn(768).reshape(shapes["0"].lens()).astype(np.float16)
-    params["0"] = migraphx.argument(arg0)
+    params["0"] = np.random.randn(768).reshape(shapes["0"].lens()).astype(
+        np.float16)
 
     r = p.run(params)[-1]
     print(r)