이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 9 Aug 2018 04:45:52 +0000 (13:45 +0900)]
Enable squeeze for rank-4 input (#2234)
- Remove input rank restriction from compilation
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 9 Aug 2018 03:01:11 +0000 (12:01 +0900)]
tflite_benchmark_model will not be formatchecked (#2232)
tflite_benchmark_model will keep original source's style
so that it could be tracked and diff'ed easily.
.FORMATCHEKED in tools (the parent directory of tflite_benchmark_model)
is distributed to all child directories except for
tflite_benchmark_model.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 9 Aug 2018 01:21:39 +0000 (10:21 +0900)]
Introduce BUILD_TFLITE_BENCHMARK_MODEL option (#2226)
Related Issue: #1760, #1696
`tflite_benchmark_model` is a tool for profiling a given model.
The tools is under development in tensorflow upstream.
It was not shipped in v1.9.0 release.
Thus I made `tflite_benchmark_model` work for tflite interpreter
by chery-picking needed patches. (#2212)
This patch is a part of full changes. It only introduces build option.
By default, tflite_benchmark_model will not be built.
To build tflite_benchmark_model tool, use
```
$ OPTIONS='-DBUILD_PURE_ARM_COMPUTE=ON -DBUILD_TFLITE_BENCHMARK_MODEL=ON' \
CROSS_BUILD=1 TARGET_ARCH=armv7l make all install
```
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 8 Aug 2018 22:18:15 +0000 (07:18 +0900)]
[neurun] Define operand usage enum and check exception (#2213)
Define operand usage: model input, constant, operation output
Check exception: operand usage conflict
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 22:16:50 +0000 (07:16 +0900)]
[neurun] Graph : Introduce operation Concat (#2223)
Introduce `neurun::graph::operation::Concat` which is a node in Graph
IR.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 22:16:39 +0000 (07:16 +0900)]
[neurun] Graph : Introduce operation Conv2D (#2222)
Introduce `neurun::graph::operation::Conv2D` which is a node in Graph
IR.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 22:16:29 +0000 (07:16 +0900)]
[neurun] Graph : Introduce operation MaxPool2D (#2220)
Introduce `neurun::graph::operation::MaxPool2D` which is a node in
Graph IR.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 22:16:16 +0000 (07:16 +0900)]
[neurun] Graph : Introduce operation Reshape (#2219)
Introduce `neurun::graph::operation::Reshape` which is a node in Graph
IR.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 22:16:05 +0000 (07:16 +0900)]
[neurun] Graph : Introduce operation FullyConnected (#2221)
Introduce `neurun::graph::operation::FullyConnected` which is a node
in Graph IR.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 22:15:49 +0000 (07:15 +0900)]
[neurun] Graph : Introduce operation Softmax (#2218)
Introduce `neurun::graph::operation::Softmax` which is a node in Graph
IR.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 11:07:23 +0000 (20:07 +0900)]
[neurun] Graph : Introduce operation `Set` (#2215)
`neurun::graph::operation::Set` holds operations and also has
ownership of them.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 10:43:36 +0000 (19:43 +0900)]
[neurun] Graph : Introduce operation AvgPool2D (#2217)
Introduce `neurun::graph::operation::AvgPool2D` which is a node in Graph
IR.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 8 Aug 2018 09:37:03 +0000 (18:37 +0900)]
support rank 4 feature in pacl runtime (#2183)
* support rank 4 feature in pacl runtime
This commit enables rank 4 feature
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
* rank4 feature for neurun
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 09:33:42 +0000 (18:33 +0900)]
[neurun] Remove kernel source from build (#2216)
SOURCES_KERNEL is an empty variable.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 08:38:57 +0000 (17:38 +0900)]
[neurun] Remove unused includes for Initializer in compilation.cc (#2214)
This commit removes unused includes for Initializer in compilation.cc
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 08:36:25 +0000 (17:36 +0900)]
[neurun] Graph : Introduce operand IndexSet (#2209)
`neurun::graph::operand::IndexSet` is a list of `Index`.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 8 Aug 2018 08:28:44 +0000 (17:28 +0900)]
Update tensorflow submodule for tflite_benchmark_model (#2212)
tflite_benchmark_model in v1.9.0 does not build nor work.
This patch fixes build failure with two cherry-picked patches.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 08:26:20 +0000 (17:26 +0900)]
[neurun] Graph : Introduce operand Set (#2208)
`neurun::graph::operand::Set` holds operands and also has ownership of
them.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 08:24:15 +0000 (17:24 +0900)]
[neurun] Graph : Introduce Operation `Node` (#2206)
Class `neurun::graph::operation::Node` is a wrapper class of
`internal::tflite::op::Node` with input/output tensor info.
Input/Output info will be used for graph traversing.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 07:54:33 +0000 (16:54 +0900)]
[neurun] Set input/output as common tensor,Implement tensor conversion kernels (#2211)
Related : #2142, #2040
- Set input/output as common tensor
- Implement tensor conversion kernels
Tests are matched.
Signed-off-by: sjsujinkim sjsujin.kim@samsung.com
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 8 Aug 2018 06:53:53 +0000 (15:53 +0900)]
Remove newruntime_frameworktest_list.txt (#2207)
Remove newruntime_frameworktest_list.txt file
It is not used any more
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 05:55:20 +0000 (14:55 +0900)]
[neurun] Change parameters for TensorConversion kernels (#2187)
Related : #2142
This commit changes parameters for TensorConversion kernels.
- Introduce Tensor Shape parameter for iterating tensors by shape info
- Change CPU Data parameter from a pointer to Tensor
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 05:07:52 +0000 (14:07 +0900)]
[neurun] Graph : Introduce Operand Index (#2205)
Introduce `neurun::graph::operand::Index` which is an alias of
`neurun::graph::Index`.
Note that, with the additional template parameter `DummyTag`, aliased
types are treated as different types(strong typedef).
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 05:07:28 +0000 (14:07 +0900)]
[neurun] Graph : Introduce Operation Index (#2204)
Introduce `neurun::graph::operation::Index` which is an alias of
`neurun::graph::Index`.
Note that, with the additional template parameter `DummyTag`, aliased
types are treated as different types(strong typedef).
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 04:13:48 +0000 (13:13 +0900)]
[neurun] Introduce Index (#2200)
Class `Index` is an integer index wrapper which will be used for
operand, operation and so on.
With template parameter `DummyTag` we can reuse code for many kinds of
indices without adding any code.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 04:13:41 +0000 (13:13 +0900)]
[neurun] Build graph sources (#2201)
Revise CMakeLists.txt to build source files in `graph` directory.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 8 Aug 2018 02:19:34 +0000 (11:19 +0900)]
[neurun] Handle more operand type (#2190)
- Handle more operand type: ANEURALNETWORKS_TENSOR_INT32, ANEURALNETWORKS_TENSOR_QUANT8_ASYMM
- Handle exception case related with type
- Update unittest skiplist
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 02:19:20 +0000 (11:19 +0900)]
[neurun] Allow to have multi objects in an one index. (#2182)
This commit allows to have multi objects in an one index.
It's for objects of heterogeneous backends in an one index.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 8 Aug 2018 02:19:04 +0000 (11:19 +0900)]
[neurun] Setup unittest infra (#2191)
Setup unittest infra with GTest with a sample test.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 7 Aug 2018 10:43:45 +0000 (19:43 +0900)]
[neurun] Use std::nothrow for exception handling (#2184)
Use std::nothrow for
- Handle exception
- Set returned reference by parameter nullptr
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 7 Aug 2018 10:43:35 +0000 (19:43 +0900)]
[neurun] Mark FromCommon for all of ifm indexes in concat (#2185)
This commit marks FromCommon for all of ifm indexes in concat.
It'was a bug.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 7 Aug 2018 08:56:27 +0000 (17:56 +0900)]
Test driver: report directory (#2188)
Add parameter to set report directory
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 7 Aug 2018 08:54:34 +0000 (17:54 +0900)]
Update unittest skiplist: validation test (#2181)
Update validation unittest skiplist
- Remove CreateModel in pacl & neurun
- Remove Model.finish in neurun
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 7 Aug 2018 01:36:45 +0000 (10:36 +0900)]
[neurun] Model finish flag (#2176)
Introduce model finish flag to check duplicated finish call & modification after finished
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 7 Aug 2018 00:42:53 +0000 (09:42 +0900)]
Introduce docker-shell and docker-run (#2177)
Change script name docker_runs.sh to docker-shell
Introduce docker-run
These scripts' name are same as nncc
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 6 Aug 2018 05:45:12 +0000 (14:45 +0900)]
[pacl] NNAPI frontend exception handling by nullptr (#2164)
Handle exception by nullptr parameter in NNAPI frontend implementation
Same implementation as neurun
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 3 Aug 2018 07:00:34 +0000 (16:00 +0900)]
[neurun] Fix convolution cpu backend debug build bug (#2150)
Fix convolution cpu backend debug build bug by check code in tflite kernel
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 3 Aug 2018 07:00:21 +0000 (16:00 +0900)]
Update TestValidation.cpp (#2161)
Update TestValication.cpp from android code
Update skiplist
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
박세희/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 3 Aug 2018 06:05:00 +0000 (15:05 +0900)]
Fix for Android build (#2160)
This will for for Android build to reflect recent changes
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 3 Aug 2018 05:51:43 +0000 (14:51 +0900)]
[neurun] Implement StageGenerator for TensorConvert nodes (#2151)
This commit implements StageGenerator for TensorConvert nodes.
Signed-off-by: sjsujinkim sjsujin.kim@samsung.com
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 3 Aug 2018 04:37:45 +0000 (13:37 +0900)]
[neurun] NNAPI frontend exception handling by nullptr (#2157)
Handle exception by nullptr parameter in NNAPI frontend implementation
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 3 Aug 2018 02:16:01 +0000 (11:16 +0900)]
[neurun] Split execution NNAPI implementation into frontend (#2152)
Split exeution NNAPI implementation into frontend
Move Sink and Source class into internal
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 2 Aug 2018 10:35:01 +0000 (19:35 +0900)]
[neurun] Bug fix: ANeuralNetworksCompilation_finish (#2153)
Remove duplicated return in ANeuralNetworksCompilation_finish
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 2 Aug 2018 07:32:44 +0000 (16:32 +0900)]
[neurun] Introduce Common Operands in Plan (#2144)
Related : #2142
This commit introduces common operands in Plan, since we should set input/output as Common Tensor.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 2 Aug 2018 07:29:37 +0000 (16:29 +0900)]
[neurun] Insert TensorConvert nodes (#2127)
* [neurun] Insert TensorConvert nodes
This commit inserts TensorConvert nodes. Those will be inserted to between all of nodes.
Signed-off-by: sjsujinkim sjsujin.kim@samsung.com
* Change to plugin type
* Fix build failure
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 2 Aug 2018 07:01:37 +0000 (16:01 +0900)]
Disable submodule update for obs build (#2149)
Submodule update is disabled in obs build
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 2 Aug 2018 05:42:02 +0000 (14:42 +0900)]
Add logging of input/output tensors of Mul_1 quicktest (#2140)
By this commit, Mul_1 quicktest logs input/output tensor values into specified log file.
The log file is used to create a new Generated File.
Related PR: #2100 (Logger),
#2105 (#2105 used log file data of this PR to create a new generated test)
Please refer to the comment in #2100 to see how to convert quicktest to generated test.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 2 Aug 2018 05:05:01 +0000 (14:05 +0900)]
Update tensorflow submodule (#2143)
Update tensorflow submodule to update nnapi_delegate and NeuralNetworksEx.h
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 2 Aug 2018 04:55:46 +0000 (13:55 +0900)]
[neurun] split compilation NNAPI implementation into frontend (#2145)
Split compilation NNAPI implementation into frontend
Define ANeuralNetworksCompilation::finish() to divide NNAPI implementation with internal
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 2 Aug 2018 04:08:24 +0000 (13:08 +0900)]
Environment variable to set all configurable operators' backend (#2141)
Introduce new environment variable to set all configurable operators' backend: OP_BACKEND_ALLOPS
none: use each operator's backend configuration (default)
acl_cl: use acl for all configurable operators' backend
cpu: use cpu for all configurable operators' backend
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 2 Aug 2018 01:42:49 +0000 (10:42 +0900)]
[neurun] change logging environment variable name (#2138)
Change logging environment variable name: NEURUN_LOG_ENABLE
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 2 Aug 2018 00:36:55 +0000 (09:36 +0900)]
Detect framework test failure (#2137)
Return exit code in run_test.sh to detect framework test error in run_frameworktest.sh
Fix bug (by additional space) in run_frameworktest.sh
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 1 Aug 2018 08:11:37 +0000 (17:11 +0900)]
Bug fix: support strided slice (#2136)
Support strided slice operator
STRIDED_SLICE_EX is removed
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 1 Aug 2018 07:41:25 +0000 (16:41 +0900)]
Move memory and model NNAPI implementation to frontend (#2133)
- Move NNAPI implementation in model.cc into frontend
- Move NNAPI implementation in memory.cc into frontend
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 1 Aug 2018 07:39:15 +0000 (16:39 +0900)]
Update NeuralNetworksEx.h (#2134)
- Remove redundant operator
- Remove unittest
- Remove implementation in runtime
- Add SPLIT operator
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 1 Aug 2018 07:02:18 +0000 (16:02 +0900)]
More `mul` tests for generated test (#2105)
1. converts quick tests for adas to generated test
2. remove successful tests from skip files
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 1 Aug 2018 02:59:33 +0000 (11:59 +0900)]
Introduce frontend directory (#2130)
- Make frontend directory
- Move event.cc into frontend
- Modify cmake to build frontend
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 1 Aug 2018 01:25:33 +0000 (10:25 +0900)]
Logging of input and output tensor of quicktest (#2100)
This is for #2047.
By providing `tensor_logging` param to `RandomTestRunner`,
input and output tensor of quicktest will be saved into specified file.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 31 Jul 2018 09:45:56 +0000 (18:45 +0900)]
[neurun] Change namespace of kernels (#2126)
This commit changes namespace of kernels from internal::kernel to neurun::kernel.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 31 Jul 2018 09:45:45 +0000 (18:45 +0900)]
[neurun] Change namespace of cpu backend (#2123)
Change namespace `internal::cpu` to `neurun::backend::cpu`.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 31 Jul 2018 07:54:59 +0000 (16:54 +0900)]
Framework test list file (#2121)
- Remove pacl framework test list: use default setting
- Define neurun framework test list: newruntime_framworktest_list.txt will be removed
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 31 Jul 2018 07:44:19 +0000 (16:44 +0900)]
[neurun] Add skeleton of tensor conversion kernels (#2065)
* [neurun] Add skeleton of tensor conversion kernels
Related : #1974
This commit adds skeleton of tensor conversion kernels inherited arm_compute::IFunction.
It's part PR of #2056(https://github.sec.samsung.net/STAR/nnfw/pull/2056#issuecomment-128722)
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
* Seperate specific layers and apply common tensor
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 31 Jul 2018 07:42:12 +0000 (16:42 +0900)]
Remove docker_env_runtime file (#2122)
Remove docker_env_runtime file: not used any more
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 31 Jul 2018 07:30:24 +0000 (16:30 +0900)]
Squeeze 2D input test (for neural_networks_test/generated) (#2087)
This commit adds Squeeze against 2D input test.
C++ files are generated by `./runtimes/tests/neural_networks_test/specs/generate_test.sh`
How to run:
`$ LD_LIBRARY_PATH=Product/out/lib Product/out/unittest/runtime_run_android_nn_test --gtest_filter=GeneratedTests.squeeze_2D*`
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 31 Jul 2018 06:55:48 +0000 (15:55 +0900)]
Add optimizeGraph to model_freezer_util (#2118)
This function calls optimize_for_inference of tensorflow and
generates '*_optimized.pb'.
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 31 Jul 2018 02:03:19 +0000 (11:03 +0900)]
Removed passed test cases from skip files (#2117)
For armv7l-linux: pevious (33 test cases) -> now (64 test cases)
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 30 Jul 2018 23:27:06 +0000 (08:27 +0900)]
removed passed CONV tests from skip list of Generated Test (#2112)
Three `CONV`-related tests were passed and removed from skip file.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 30 Jul 2018 10:48:50 +0000 (19:48 +0900)]
[neurun] Change namespace for acl_cl backend (#2106)
Change namespace `internal::arm_compute` to `neurun::backend::acl_cl`.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
남궁석/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 30 Jul 2018 08:03:06 +0000 (17:03 +0900)]
[neurun] Update Inputcount Restriction (#2108)
Update ANEURALNETWORKS_CONV_2D explicit inputCount from 9 to 10
Update ANEURALNETWORKS_MAX_POOL_2D explicit inputCount from 9 to 10
Update ANEURALNETWORKS_AVERAGE_POOL_2D explicit inputCount from 9 to 10
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 30 Jul 2018 08:02:20 +0000 (17:02 +0900)]
[neurun] Update library info string (#2107)
Update library info string `new` to `neurun`.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 30 Jul 2018 06:59:12 +0000 (15:59 +0900)]
[neurun] Add Common Tensor/TensorBuilder (#2082)
* [neurun] Add Common Tensor/TensorBuilder
This commit adds Common Tensor/TensorBuilder for converting between backends.
The name of 'Common' would be changed or kept.
* Remove common tensor info context
* Replace common_tensor_builder to BackendManager
남궁석/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 30 Jul 2018 04:27:49 +0000 (13:27 +0900)]
Add python script of run_unittest (#2050)
Parent Issue : #1896
Working Issue : #1019
This commit will add python script of run_unittest
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 30 Jul 2018 04:19:36 +0000 (13:19 +0900)]
[model_parser] Read BuiltinOptions dynamically (#2094)
This commit removes the hardcoded implementation of BuiltinOptions
Generator table. It now reads the list of Builtin Options from
`BuiltinOptions.py` and automatically imports the corresponding
classes and build the generator map dynamically. All these were
possible with reflection.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
남궁석/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 30 Jul 2018 03:48:32 +0000 (12:48 +0900)]
[neurun] Update Library Information Name (#2102)
Apply our new runtime name `neurun` to library information name
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 27 Jul 2018 04:27:55 +0000 (13:27 +0900)]
[neurun] Build acl_cl kernel as a separate lib (#2095)
Build acl_cl kernel as a separate static lib `libkernel_acl_cl.a`.
Likewise #2084 for cpu kernel.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 26 Jul 2018 09:57:40 +0000 (18:57 +0900)]
ACL patch for mean operation (#2093)
This commit introduce ACL patch for mean operation.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 26 Jul 2018 08:24:08 +0000 (17:24 +0900)]
[model_parser] Print activation function as name (#2092)
Print Activation Function Type as its name, not the integer code.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 26 Jul 2018 08:20:19 +0000 (17:20 +0900)]
Removed GeneratedTests.squeeze_float_1 from test skip list (#2089)
`GeneratedTests.squeeze_float_1` works so removed from test skip list.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 26 Jul 2018 04:57:21 +0000 (13:57 +0900)]
[model_parser] Generalize `SetBuiltinOpcodeStr` (#2073)
Generalize the function `SetBuiltinOpcodeStr` so it can be used from
other enum classes as well. Plus, introduce `EnumStrMaps` which is a
collection of these kind of map(enum to string conversion).
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 26 Jul 2018 04:46:30 +0000 (13:46 +0900)]
[neurun] Build cpu kernel as separate lib (#2084)
Build cpu kernel as separate lib `libkernel_cpu.a`.
Part of #1960
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 26 Jul 2018 02:04:36 +0000 (11:04 +0900)]
[neurun] Introduce tensor conversion Nodes (#2064)
* [neurun] Introduce tensor conversion Nodes
Related : #1974
This commit introduces tensor conversion Nodes.
- FromCommon : "common order shape" -> " specific shape of next op".
- ToCommon : "specific shape of prev op" -> "common shape".
The name of 'Common' would be changed to another.
It's part PR of #2056(https://github.sec.samsung.net/STAR/nnfw/pull/2056#issuecomment-128722)
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
* Remove unnecessary op list
* Seperate ACL/CPU converter Nodes
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 26 Jul 2018 01:59:33 +0000 (10:59 +0900)]
Introduce addShapeConstr considering broadcasting for ADD, SUB, DIV, MUL (#1852)
* Introduce addShapeConstr considering broadcasting
This commit introduces addShapeConstr considering broadcasting for ADD, SUB, DIV, MUL
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Delete useless cout statement
This commit deletes useless cout statement
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Modify description and typos
This commit modifies description and typos.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 26 Jul 2018 01:59:25 +0000 (10:59 +0900)]
Introduce ADD broadcast for inputs without initialized data (#1780)
This commit introduces ADD broadcast for inputs without initialized data
-In case input data is not initialized.
-related with #1765 (inputs with initialized data)
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 26 Jul 2018 01:54:48 +0000 (10:54 +0900)]
Change to use RandomGenerator outside the RandomTestRunner. (#1922)
This commit changes to use RandomGenerator outside of RandomTestRunner because some inputs are able to set the value outside of RandomTestRunner.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 26 Jul 2018 01:37:09 +0000 (10:37 +0900)]
Not disable all test when setting framework test list (#2081)
Not disable all test when setting framework test list
It will be used for neurun test
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
남궁석/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 26 Jul 2018 00:20:26 +0000 (09:20 +0900)]
Enable Generated.tanh_ test for armv7l-linux (#2045)
Generated.tanh_ test in armv7l-linux was passed (#1993)
Therefore this commit will enable the test
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 25 Jul 2018 22:05:23 +0000 (07:05 +0900)]
[Pure ACL] Squeeze operation (#2038)
This commit introduces Squeeze operation. 2D, 3D are tested with generated test.
Note that 4D test (squeeze.mod.py) does not work. 4D will be worked as a separate PR.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 25 Jul 2018 10:54:20 +0000 (19:54 +0900)]
Remove unused scripts and change script name (#2083)
Remove unused script: pureacl
Change name: neurun
Remain docker_env_newruntime: CI issue
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 25 Jul 2018 09:36:24 +0000 (18:36 +0900)]
[neurun] Do not compile cpu backend from runtime (#2080)
Do not compile cpu backend sources when building runtime.
These sources was supposed to be excluded when cpu backend is built
separately but it wasn't yet.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 25 Jul 2018 08:34:12 +0000 (17:34 +0900)]
[model_parser] Change `GetBuiltinOptions` usage (#2072)
* [model_parser] Rename `GetBuiltinOptions` and usage
Rename `GetBuiltinOptions` to `GetOptions` and change the arguments.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
* Revert renaming
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 25 Jul 2018 07:02:29 +0000 (16:02 +0900)]
Unittest list (#2076)
- Remove unittest skiplist for legacy runtime
- New unittest skiplit for neurun
- Fix cmake to install skiplist pacl or neurun
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 25 Jul 2018 07:00:45 +0000 (16:00 +0900)]
[neurun] Build backends as static lib (#2074)
To reduce packaging maintanance change backend build as static.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 25 Jul 2018 01:49:40 +0000 (10:49 +0900)]
[Model Parser] Prepare argument (#2068)
Prepare argument input
- verbose level (-v, -vv)
- tensor list (-t ...)
- operator list (-o ...)
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 25 Jul 2018 01:35:08 +0000 (10:35 +0900)]
[neurun] Move `kernel` to upper directory (#2071)
Move `kernel` directory to be the top level.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 24 Jul 2018 09:33:03 +0000 (18:33 +0900)]
[neurun] Change install directory (#2070)
Change install directory name: lib/new_runtime -> lib/neurun
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 24 Jul 2018 06:15:16 +0000 (15:15 +0900)]
Revise computing instructions of operator on model parser (#2060)
Revise computing instructions of operator on model parser.
To do it, intrudece Operation class which has each instruction members
and _OperationComputeMethod, internal class, which compute for each
operator type.
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 24 Jul 2018 06:11:21 +0000 (15:11 +0900)]
Define build option for nnapi_quickcheck (#2067)
Define build option for nnapi_quickcheck
Set default value as off (not build nnapi_quickcheck)
To build nnapi_quickcheck, set OPTIONS="-DBUILD_NNAPI_QUICKCHECK=on"
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 24 Jul 2018 05:43:18 +0000 (14:43 +0900)]
Update CrossBuildForArm.md (#2063)
- `ROOTFS_ARM`(or `ROOTFS_ARM`) is not necessary to set if you installed
rootfs in default path.
- Explanation for installing cross toolchain is modified
Related Issue: #2035, #2048
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 24 Jul 2018 04:28:48 +0000 (13:28 +0900)]
Rename nnapi_unittest (#2046)
Rename nnapi_unittest to nnapi_quickcheck to avoid confusing
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 24 Jul 2018 00:38:41 +0000 (09:38 +0900)]
Migration to tensorflow v1.9.0 (#2054)
Fix test using tensorflow function
Update nnapi_delegate.cpp/h
Update supporting custom operation
Add submodules for dependency
To use this, need to update tensorflow submodule
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>