inference-engine/thirdparty/clDNN/src/arg_max_min.cpp

   1 /*
   2 // Copyright (c) 2018 Intel Corporation
   3 //
   4 // Licensed under the Apache License, Version 2.0 (the "License");
   5 // you may not use this file except in compliance with the License.
   6 // You may obtain a copy of the License at
   7 //
   8 //      http://www.apache.org/licenses/LICENSE-2.0
   9 //
  10 // Unless required by applicable law or agreed to in writing, software
  11 // distributed under the License is distributed on an "AS IS" BASIS,
  12 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13 // See the License for the specific language governing permissions and
  14 // limitations under the License.
  15 */
  16
  17 ///////////////////////////////////////////////////////////////////////////////////////////////////
  18 #include "arg_max_min_inst.h"
  19 #include "primitive_type_base.h"
  20 #include "sliding_window_utils.h"
  21 #include "error_handler.h"
  22 #include "json_object.h"
  23
  24 namespace cldnn
  25 {
  26         primitive_type_id arg_max_min_type_id()
  27         {
  28                 static primitive_type_base<arg_max_min> instance;
  29                 return &instance;
  30         }
  31
  32         layout arg_max_min_inst::calc_output_layout(arg_max_min_node const& node)
  33         {
  34         assert((bool)node.get_primitive()->output_data_type == false
  35                && "Output data type forcing is not supported for "
  36                   "arg_max_min_node!");
  37         auto desc = node.get_primitive();
  38
  39                 auto input_layout = node.input().get_output_layout();
  40
  41                 if (desc->with_axis){
  42                         switch (desc->axis){
  43                                 case arg_max_min::x:
  44                                         return layout{ data_types::f32, format::bfyx, tensor{ input_layout.size.batch[0], input_layout.size.feature[0], (int32_t)desc->top_k, input_layout.size.spatial[1] }};
  45                                 case arg_max_min::y:
  46                                         return layout{ data_types::f32, format::bfyx, tensor{ input_layout.size.batch[0], input_layout.size.feature[0], input_layout.size.spatial[0], (int32_t)desc->top_k }};
  47                                 case arg_max_min::feature:
  48                                         return layout{ data_types::f32, format::bfyx, tensor{ input_layout.size.batch[0], (int32_t)desc->top_k, input_layout.size.spatial[0], input_layout.size.spatial[1] }};
  49                                 case arg_max_min::batch:
  50                                         return layout{ data_types::f32, format::bfyx, tensor{ (int32_t)desc->top_k, input_layout.size.feature[0], input_layout.size.spatial[0], input_layout.size.spatial[1] }};
  51                                 default:
  52                                         break;
  53                         }
  54                 }
  55
  56                 return layout{ data_types::f32, input_layout.format, tensor{ input_layout.size.batch[0], 1, (int32_t)desc->top_k, 1 } };
  57         }
  58
  59         std::string arg_max_min_inst::to_string(arg_max_min_node const& node)
  60         {
  61                 auto desc = node.get_primitive();
  62                 auto node_info = node.desc_to_json();
  63                 auto axis = desc->with_axis ? "true" : "false";
  64                 auto out_type = desc->output_type ? "max" : "min";
  65
  66                 std::stringstream primitive_description;
  67
  68                 json_composite conv_info;
  69                 conv_info.add("top_k", desc->top_k);
  70                 conv_info.add("with axis", axis);
  71                 if (desc->with_axis)
  72                         conv_info.add("axis", desc->axis);
  73                 conv_info.add("output type", out_type);
  74                 node_info->add("arg_max_min info", conv_info);
  75                 node_info->dump(primitive_description);
  76
  77                 return primitive_description.str();
  78         }
  79
  80         arg_max_min_inst::typed_primitive_inst(network_impl& network, arg_max_min_node const& node)
  81                 : parent(network, node)
  82         {
  83         }
  84 }