Merge branch 'dyn_nms' of github.com:ROCmSoftwarePlatform/AMDMIGraphX into dyn_topK

src/include/migraphx/op/scatter_none.hpp

@@ -25,16 +25,8 @@
 #define MIGRAPHX_GUARD_OPERATORS_SCATTER_NONE_HPP
 
 #include <array>
-#include <migraphx/check_shapes.hpp>
-#include <migraphx/stringutils.hpp>
-#include <migraphx/streamutils.hpp>
-#include <migraphx/shape_for_each.hpp>
 #include <migraphx/config.hpp>
-#include <migraphx/value.hpp>
-#include <migraphx/op/normalize_attribute.hpp>
 #include <migraphx/op/scatter.hpp>
-#include <cmath>
-#include <utility>
 
 // Scatter op. with "none" as the reduction function (just copies the value).  This is identical to
 // the previously existing Scatter op.

src/include/migraphx/op/scatternd_op.hpp

@@ -26,6 +26,7 @@
 
 #include <migraphx/op/name.hpp>
 #include <migraphx/check_shapes.hpp>
+#include <migraphx/argument.hpp>
 #include <migraphx/par_for.hpp>
 
 namespace migraphx {

src/include/migraphx/op/sinh.hpp

@@ -24,16 +24,9 @@
 #ifndef MIGRAPHX_GUARD_OPERATORS_SINH_HPP
 #define MIGRAPHX_GUARD_OPERATORS_SINH_HPP
 
-#include <array>
 #include <migraphx/op/unary.hpp>
-#include <migraphx/check_shapes.hpp>
-#include <migraphx/stringutils.hpp>
-#include <migraphx/streamutils.hpp>
-#include <migraphx/literal.hpp>
-#include <migraphx/shape_for_each.hpp>
 #include <migraphx/config.hpp>
 #include <cmath>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {

src/include/migraphx/op/topk.hpp

@@ -26,6 +26,7 @@
 
 #include <algorithm>
 #include <migraphx/check_shapes.hpp>
+#include <migraphx/argument.hpp>
 #include <migraphx/config.hpp>
 #include <migraphx/op/normalize_attribute.hpp>
 #include <migraphx/par_for.hpp>

src/include/migraphx/op/unknown.hpp

@@ -25,7 +25,6 @@
 #define MIGRAPHX_GUARD_RTGLIB_UNKNOWN_HPP
 
 #include <migraphx/config.hpp>
-#include <migraphx/argument.hpp>
 #include <migraphx/check_shapes.hpp>
 
 namespace migraphx {

src/onnx/onnx.cpp

@@ -44,13 +44,21 @@ program parse_onnx_from(const onnx_options& options, Ts&&... xs)
     parser.map_input_dims     = options.map_input_dims;
     parser.map_dyn_input_dims = options.map_dyn_input_dims;
     auto dim_val              = options.default_dim_value;
-    if(dim_val == 0)
+    if(dim_val != 0)
     {
-        parser.default_dyn_dim_value = options.default_dyn_dim_value;
+        if(options.default_dyn_dim_value != shape::dynamic_dimension{1, 1, 0})
+        {
+            MIGRAPHX_THROW("PARSE_ONNX_FROM: both default_dim_value and default_dyn_dim_value"
+                           "set to non-default value");
+        }
+        else
+        {
+            parser.default_dyn_dim_value = {dim_val, dim_val, 0};
+        }
     }
     else
     {
-        parser.default_dyn_dim_value = {dim_val, dim_val, 0};
+        parser.default_dyn_dim_value = options.default_dyn_dim_value;
     }
     parser.skip_unknown_operators = options.skip_unknown_operators;
     parser.max_loop_iterations    = options.max_loop_iterations;

src/onnx/onnx_parser.cpp

@@ -257,6 +257,11 @@ int64_t onnx_parser::get_opset_version(const onnx::ModelProto& model)
 
 void onnx_parser::parse_graph(module* mod, const onnx::GraphProto& graph)
 {
+    if(not map_input_dims.empty() and not map_dyn_input_dims.empty())
+    {
+        MIGRAPHX_THROW("PARSE_GRAPH: both map_input_dims and map_dyn_input_dims non-empty, only"
+                       "one should be used");
+    }
     std::unordered_map<std::string, instruction_ref> mod_insts;
     for(auto&& f : graph.initializer())
     {

src/targets/gpu/include/migraphx/gpu/int8_conv_pack.hpp

@@ -24,6 +24,7 @@
 #ifndef MIGRAPHX_GUARD_RTGLIB_INT8_CONV_PACK_HPP
 #define MIGRAPHX_GUARD_RTGLIB_INT8_CONV_PACK_HPP
 
+#include <migraphx/argument.hpp>
 #include <migraphx/op/quant_dot.hpp>
 #include <migraphx/config.hpp>
 #include <utility>

src/targets/gpu/include/migraphx/gpu/int8_gemm_pack.hpp

@@ -24,6 +24,7 @@
 #ifndef MIGRAPHX_GUARD_RTGLIB_INT8_GEMM_PACK_HPP
 #define MIGRAPHX_GUARD_RTGLIB_INT8_GEMM_PACK_HPP
 
+#include <migraphx/argument.hpp>
 #include <migraphx/op/quant_dot.hpp>
 #include <migraphx/config.hpp>
 #include <utility>

src/tf/tf_parser.cpp

@@ -216,7 +216,7 @@ static std::vector<T> get_data_vals(const google::protobuf::RepeatedField<T>& da
         std::fill(data_vals.begin(), data_vals.end(), data[0]);
     }
     else
-        copy(data.begin(), data.end(), std::back_inserter(data_vals));
+        copy(data.begin(), data.end(), data_vals.begin());
     return data_vals;
 }
 
@@ -329,33 +329,37 @@ void tf_parser::parse_node(const std::string& name)
         auto&& node = nodes.at(name);
         if(not is_valid_op(node))
             return;
-
         std::vector<instruction_ref> args;
-
         for(auto&& input : node.input())
         {
             // control dependencies (signified by ^ before the name) are ignored
             if(contains(input, "^"))
                 continue;
-            if(nodes.count(input) > 0)
+            std::string input_name = input;
+            // if input has trailing `:0` index then remove it
+            auto multi_out_idx = input.find(':');
+            if(multi_out_idx != std::string::npos && input.substr(multi_out_idx + 1) == "0")
+            {
+                input_name = input.substr(0, multi_out_idx);
+            }
+            if(nodes.count(input_name) > 0)
             {
-                std::string iname;
                 // input was from a node with multiple outputs
-                if(contains(input, ':'))
+                if(contains(input_name, ':'))
                 {
-                    iname = input.substr(0, input.find(':'));
+                    input_name = input_name.substr(0, input.find(':'));
                 }
                 else
                 {
-                    iname = get_name(nodes.at(input));
+                    input_name = get_name(nodes.at(input_name));
                 }
-                assert(name != iname);
-                this->parse_node(iname);
-                args.push_back(instructions.at(input));
+                assert(name != input_name);
+                this->parse_node(input_name);
+                args.push_back(instructions.at(input_name));
             }
             else
             {
-                args.push_back(instructions.at(input));
+                args.push_back(instructions.at(input_name));
             }
         }
         std::vector<instruction_ref> result;

test/onnx/gen_onnx.py

@@ -3109,7 +3109,7 @@ def max_test():
     a = helper.make_tensor_value_info('0', TensorProto.FLOAT, [3])
     b = helper.make_tensor_value_info('1', TensorProto.FLOAT, [3])
     c = helper.make_tensor_value_info('2', TensorProto.FLOAT, [3])
-    y = helper.make_tensor_value_info('2', TensorProto.FLOAT, [3])
+    y = helper.make_tensor_value_info('3', TensorProto.FLOAT, [3])
 
     node = onnx.helper.make_node(
         'Max',
@@ -3243,7 +3243,7 @@ def min_test():
     a = helper.make_tensor_value_info('0', TensorProto.FLOAT, [3])
     b = helper.make_tensor_value_info('1', TensorProto.FLOAT, [3])
     c = helper.make_tensor_value_info('2', TensorProto.FLOAT, [3])
-    y = helper.make_tensor_value_info('2', TensorProto.FLOAT, [3])
+    y = helper.make_tensor_value_info('3', TensorProto.FLOAT, [3])
 
     node = onnx.helper.make_node(
         'Min',

test/onnx/max_test.onnx

-max-example:e
+max_test:a
 
 
0
 
1
-
2
3"Maxtest-dropoutZ
+
2
3"Maxmax_testZ
 
0
 
 

@@ -15,7 +15,7 @@
 
 

 b
-
2
+
3
 
 

-B
\ No newline at end of file
+B
\ No newline at end of file

test/onnx/min_test.onnx

-min-example:e
+min_test:a
 
 
0
 
1
-
2
3"Mintest-dropoutZ
+
2
3"Minmin_testZ
 
0
 
 

@@ -15,7 +15,7 @@
 
 

 b
-
2
+
3
 
 

-B
\ No newline at end of file
+B
\ No newline at end of file

test/onnx/onnx_test.cpp

@@ -2874,7 +2874,9 @@ TEST_CASE(max_test)
     auto l0     = mm->add_instruction(migraphx::make_op("max"), input0, input1);
     mm->add_instruction(migraphx::make_op("max"), l0, input2);
 
-    optimize_onnx("max_test.onnx");
+    auto prog = optimize_onnx("max_test.onnx");
+
+    EXPECT(p == prog);
 }
 
 TEST_CASE(maxpool_notset_test)
@@ -2989,7 +2991,9 @@ TEST_CASE(min_test)
     auto l0     = mm->add_instruction(migraphx::make_op("min"), input0, input1);
     mm->add_instruction(migraphx::make_op("min"), l0, input2);
 
-    optimize_onnx("min_test.onnx");
+    auto prog = optimize_onnx("min_test.onnx");
+
+    EXPECT(p == prog);
 }
 
 TEST_CASE(multinomial_test)
@@ -5573,6 +5577,26 @@ TEST_CASE(variable_batch_user_input_test4)
     EXPECT(p == prog);
 }
 
+TEST_CASE(variable_batch_user_input_test5)
+{
+    // Error using default_dim_value and default_dyn_dim_value
+    migraphx::onnx_options options;
+    options.default_dim_value     = 2;
+    options.default_dyn_dim_value = {1, 2, 0};
+
+    EXPECT(test::throws([&] { migraphx::parse_onnx("variable_batch_test.onnx", options); }));
+}
+
+TEST_CASE(variable_batch_user_input_test6)
+{
+    // Error using both map_dyn_input_dims and map_input_dims
+    migraphx::onnx_options options;
+    options.map_dyn_input_dims["0"] = {{2, 5, 0}, {3, 3, 0}, {16, 16, 0}, {16, 16, 0}};
+    options.map_input_dims["0"]     = {2, 3, 16, 16};
+
+    EXPECT(test::throws([&] { migraphx::parse_onnx("variable_batch_test.onnx", options); }));
+}
+
 TEST_CASE(variable_batch_leq_zero_test)
 {
     migraphx::program p;