Merge branch 'lenet-test' of https://github.com/ROCmSoftwarePlatform/RTGLib into lenet-test

src/onnx/CMakeLists.txt

@@ -8,4 +8,4 @@ target_link_libraries(onnx-proto PRIVATE ${PROTOBUF_LIBRARY})
 
 add_executable(read_onnx read_onnx.cpp)
 rocm_clang_tidy_check(read_onnx)
-target_link_libraries(read_onnx onnx-proto rtg)
+target_link_libraries(read_onnx onnx-proto rtg rtg_cpu)

src/onnx/read_onnx.cpp

@@ -12,6 +12,9 @@
 #include <rtg/program.hpp>
 #include <rtg/operators.hpp>
 
+#include <rtg/cpu/cpu_target.hpp>
+#include <random>
+
 struct unknown
 {
     std::string op;
@@ -227,6 +230,13 @@ struct onnx_parser
         return result;
     }
 
+    template <class T>
+    static rtg::literal from_repeated(rtg::shape::type_t t, const T& r)
+    {
+        std::size_t size = r.size();
+        return rtg::literal{{t, {size}}, r.begin(), r.end()};
+    }
+
     static rtg::literal parse_value(const onnx::AttributeProto& attr)
     {
         switch(attr.type())
@@ -238,10 +248,8 @@ struct onnx_parser
         case onnx::AttributeProto::TENSOR: return parse_tensor(attr.t());
         case onnx::AttributeProto::GRAPH: return {};
         case onnx::AttributeProto::FLOATS:
-            return rtg::literal{rtg::shape::float_type, attr.floats().begin(), attr.floats().end()};
-        case onnx::AttributeProto::INTS:
-            return rtg::literal{rtg::shape::int32_type, attr.ints().begin(), attr.ints().end()};
-            ;
+            return from_repeated(rtg::shape::float_type, attr.floats());
+        case onnx::AttributeProto::INTS: return from_repeated(rtg::shape::int64_type, attr.ints());
         case onnx::AttributeProto::STRINGS: return {};
         case onnx::AttributeProto::TENSORS: return {};
         case onnx::AttributeProto::GRAPHS: return {};
@@ -329,6 +337,22 @@ struct onnx_parser
     }
 };
 
+// TODO: Move this to a seperate header
+std::vector<float> get_tensor_data(rtg::shape s)
+{
+    std::vector<float> result(s.elements());
+    std::mt19937 engine{0};
+    std::uniform_real_distribution<> dist;
+    std::generate(result.begin(), result.end(), [&] { return dist(engine); });
+    return result;
+}
+
+rtg::argument get_tensor_argument(rtg::shape s)
+{
+    auto v = get_tensor_data(s);
+    return {s, [v]() mutable { return reinterpret_cast<char*>(v.data()); }};
+}
+
 int main(int argc, char const* argv[])
 {
     if(argc > 1)
@@ -339,6 +363,11 @@ int main(int argc, char const* argv[])
         try
         {
             parser.parse_from(input);
+            parser.prog.compile(rtg::cpu::cpu_target{});
+            auto s      = parser.prog.get_parameter_shape("Input3");
+            auto input3 = get_tensor_argument(s);
+            auto out    = parser.prog.eval({{"Input3", input3}});
+            (void)out;
         }
         catch(...)
         {

src/targets/cpu/cpu_target.cpp

@@ -586,6 +586,10 @@ struct cpu_apply
             {
                 apply_activation(it);
             }
+            else if(it->op.name() == "pooling")
+            {
+                apply_pooling(it);
+            }
             else if(apply_map.count(it->op.name()) > 0)
             {
                 apply_map.at(it->op.name())(it);