[manager] disable inference_memory_opt

Enable inference_inout_memory_opt in inference could make a problem

Cause: If the networks are not flatten which means have a cycle, some of output of layer should be preserved.
For example, in resnet there is a residual output and if the opt is enabled this data could be corrupted.

See Also #1094

Self evaluation:

Build test: [X]Passed [ ]Failed [ ]Skipped
Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>

diff --git a/nntrainer/tensor/manager.h b/nntrainer/tensor/manager.h
index ae7a30a60f41a3d882e905980d97128763edbbb4..85ae36c2c82ea65e18e539e599998acf34e4ee1f 100644 (file)
--- a/nntrainer/tensor/manager.h
+++ b/nntrainer/tensor/manager.h
@@ -93,7 +93,7 @@ public:
   Manager(bool enable_gradient_memory_opt_ = true,
           bool enable_derivative_memory_opt_ = true,
           bool enable_activation_memory_opt_ = true,
-          bool enable_inference_inout_memory_opt_ = true);
+          bool enable_inference_inout_memory_opt_ = false);
 
   Manager(const Manager &) = default;