[exo-tflite] Use shared shape inference framework by default (#6390)
author박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 <jh1302.park@samsung.com>
Thu, 8 Aug 2019 07:41:42 +0000 (16:41 +0900)
committerGitHub Enterprise <noreply-CODE@samsung.com>
Thu, 8 Aug 2019 07:41:42 +0000 (16:41 +0900)
This commit makes exo-tflite to use the shape inference framework
implemented in loco core by default.

Please note that it is possible to turn off this feature via setting
EXOTFLITE_USE_LOCO_SHAPE_INFERENCE as 0.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
compiler/exo-tflite/src/ShapeInference.cpp

index f14486d..ede03d5 100644 (file)
@@ -38,9 +38,9 @@ struct Knob
 {
   Knob()
   {
-    // Off by default
+    // On by default
     auto s = std::getenv("EXOTFLITE_USE_LOCO_SHAPE_INFERENCE");
-    enable_loco_shape_inferene_framework = pepper::safe_strcast<int>(s, 0 /* DEFAULT */) != 0;
+    enable_loco_shape_inferene_framework = pepper::safe_strcast<int>(s, 1 /* DEFAULT */) != 0;
   }
 
   bool enable_loco_shape_inferene_framework = false;