[Layer] Update forward and backward layer functions
authorParichay Kapoor <pk.kapoor@samsung.com>
Mon, 3 Aug 2020 09:08:02 +0000 (18:08 +0900)
committerJijoong Moon <jijoong.moon@samsung.com>
Tue, 4 Aug 2020 02:13:41 +0000 (11:13 +0900)
commit236f4d889c9f04020cf9583d979b8118ef57c11d
tree8aee0b5bd35339dd60790989884c46615d036701
parentab4ee0fb950931e1b499aa26ed2a252c764224b7
[Layer] Update forward and backward layer functions

Update the forwarding and backwarding layer functions to support
multiple inputs and outputs.
This is crucial to support concat, element wise add, etc kind of layers
as well as skip connection.
Further, forward no longer returns status but rather throws.

**Self evaluation:**
1. Build test: [x]Passed [ ]Failed [ ]Skipped
2. Run test: [x]Passed [ ]Failed [ ]Skipped

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
25 files changed:
Applications/Classification/jni/main.cpp
Applications/LogisticRegression/jni/main.cpp
Applications/ReinforcementLearning/DeepQ/jni/main.cpp
Applications/Training/jni/main.cpp
nntrainer/include/activation_layer.h
nntrainer/include/bn_layer.h
nntrainer/include/conv2d_layer.h
nntrainer/include/fc_layer.h
nntrainer/include/flatten_layer.h
nntrainer/include/input_layer.h
nntrainer/include/layer.h
nntrainer/include/loss_layer.h
nntrainer/include/neuralnet.h
nntrainer/include/pooling2d_layer.h
nntrainer/include/tensor.h
nntrainer/src/activation_layer.cpp
nntrainer/src/bn_layer.cpp
nntrainer/src/conv2d_layer.cpp
nntrainer/src/fc_layer.cpp
nntrainer/src/flatten_layer.cpp
nntrainer/src/input_layer.cpp
nntrainer/src/loss_layer.cpp
nntrainer/src/neuralnet.cpp
nntrainer/src/pooling2d_layer.cpp
test/unittest/unittest_nntrainer_layers.cpp