tflite_benchmark: use get_env_bool instead of getenv (#1812)
* tflite_benchmark: use get_env_bool instead of getenv
Replace std::getenv by nnfw::util::get_env_bool for USE_NNAPI
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
* Update tflite_benchmark.cc
tflite_benchmark: get env refactor (#1801)
* tflite_benchmark: get env refactor
using nnfw::util::get_env_int instead of custom get_env call
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
* add const spec
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
* fix compare signed/unsigned issue
tflite_benchmark: use get_env_int for THREAD (#1840)
Replace std::getenv by nnfw::util::get_env_int for THREAD
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 3 Jul 2018 10:34:36 +0000 (19:34 +0900)]
Consider different rank between left and right operands for broadcasting (#1786)
This commit resolve different rank between left and right operands for broadcasting
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
fix get_env_bool for "true"/"false" values (#1811)
* fix get_env_bool for "true"/"false" values
exception in case non int string passed
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
* remove unused code
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
남궁석/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 3 Jul 2018 05:28:50 +0000 (14:28 +0900)]
Modify Framework test script (#1829)
Setting of NNFW_LD_LIBRARY_PATH is done many times
Even though LD_LIBRARY_PATH is given, script search NNFW_LD_LIBRARY_PATH first.
Put the codes of setting NNFW_LD_LIBRARY_PATH outside of iteration block
Change the sequence of LD_LIBRARY_PATH and NNFW_LD_LIBRARY_PATH
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 3 Jul 2018 02:53:18 +0000 (11:53 +0900)]
Separate make_default_quantization function. (#1823)
This commit seperate make_default_quantization to avoid error of redefinition.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 3 Jul 2018 00:59:07 +0000 (09:59 +0900)]
Move ops_counters code to op_counter.py (#1813)
* Move ops_counters code to op_counter.py
Move ops_counters code to op_counter.py
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
* Add code to prevent no key sitatuion
* Add operator types for inceptionv3
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 2 Jul 2018 07:35:42 +0000 (16:35 +0900)]
Update NeuralNetworks.h: NN API 1.1 (#1815)
* Update NeuralNetworks.h: NN API 1.1
Update NeuralNetworks.h to released NN API 1.1
https://android.googlesource.com/platform/frameworks/ml/+/android-p-preview-3/nn/runtime/include/NeuralNetworks.h
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Remove andorid api version check
Remove andorid api version check
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Remove base version check
Remove base android version check
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 2 Jul 2018 07:11:44 +0000 (16:11 +0900)]
Do not install nnapi_unittest testcases (#1817)
These testcases are not an official test tools
So do not install nnapi_unittest testcases to unitteset directory.
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
남궁석/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 2 Jul 2018 02:26:44 +0000 (11:26 +0900)]
Updates for cross build using docker (#1807)
Modify some sentences
Add method about cross build using docker with ACL build
Remove method for cross build using docker without ACL build
(We don't use method without ACL build anymore)
env: add default value parameter to get_env (#1800)
* env: add default value parameter to get_env
defaultValue parameter can be used if named parametr is missing
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
* move default value to decaration
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 2 Jul 2018 01:50:10 +0000 (10:50 +0900)]
Remove dependency with internal header for custom op (#1804)
Remove dependency with internal header to support custom operation
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 29 Jun 2018 08:19:37 +0000 (17:19 +0900)]
Adds counter for operations of convolution operator (#1803)
* Adds counter for operations of convolution operator
Adds counter for operations of convolution operator
- Ref:
https://github.com/apaszke/torch-opCounter/blob/
e610cebf12c7ba36f5ee1c6b327129478202b32e/src/profiler.lua#L75
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
* Apply format.patch
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 29 Jun 2018 04:04:07 +0000 (13:04 +0900)]
Add nnapi_unittest for ReLU6 (2D-tensor) (#1794)
relu6_1 is for 2 ranks input model.
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 29 Jun 2018 00:49:40 +0000 (09:49 +0900)]
Update acl patch (gather operation) (#1769)
This commit updates acl patch.
-https://github.sec.samsung.net/RS7-RuntimeNTools/ComputeLibrary/pull/26
-related with #1746
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 28 Jun 2018 07:47:11 +0000 (16:47 +0900)]
Add an unit test for TANH op (#1774)
This commit adds an unit test for TANH op.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 28 Jun 2018 06:30:09 +0000 (15:30 +0900)]
Use GLOB in nn runtime's CMakeLists.txt (#1798)
Use GLOB for listing all sources becasue ...
- Currently we compile all *.cpp sources.
- By using GLOB, we don't need to modify CMakeLists.txt.
- Our other runtimes (pure_arm_compute, new_runtime) already use this feature.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 28 Jun 2018 06:29:27 +0000 (15:29 +0900)]
Support ReLU1 operation (#1795)
Define ReLU1 namespace, node, param in pure acl runtime
Set parameter for ReLU1 in pure acl runtime
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 28 Jun 2018 03:36:57 +0000 (12:36 +0900)]
Add runtime information in libneuralnetworks.so (#1793)
Related Issue: #1792
You can get the name of runtime from `libneuralnetworks.so` by command:
```
$ strings Product/out/lib/libneuralnetworks.so | grep "library info"
library information : runtime=nn
$ strings Product/out/lib/pureacl/libneuralnetworks.so | grep "library info"
library information : runtime=pure_arm_compute
```
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 28 Jun 2018 01:46:58 +0000 (10:46 +0900)]
Support ReLU6 operation (#1770)
Define ReLU6 namespace, node, param in pure acl runtime
Set parameter for ReLU6 in pure acl runtime
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
tflite_benchmark: Fix segfault on model loading. (#1782)
* tflite_benchmark: Fix segfault on model loading.
add BuildFromFile return result checking
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
* cout to cerr
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
Add tflite_benchmark console params checking (#1781)
* Add tflite_benchmark console params checking
add params count checking
add help
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
* add ostream as param
Signed-off-by: Vitaliy Cherepanov <v.cherepanov@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 27 Jun 2018 10:44:31 +0000 (19:44 +0900)]
Add testcases for SUB broadcasting (#1789)
This commit adds testcases for SUB broadcasting
-left > right (rank): test5
-left < right (rank): test6
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 27 Jun 2018 09:25:20 +0000 (18:25 +0900)]
Print operator types and its count for model (#1790)
Print operator types and its count for model
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 27 Jun 2018 02:38:29 +0000 (11:38 +0900)]
Change model_parser.py to print tensors of each input/output (#1785)
* Change model_parser.py to print tensors of each input/output
Change model_parser.py to print tensors of each input/output. This
commit makes users easily see input/output tensors.
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
* fix format to pass format-checker
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 27 Jun 2018 01:02:03 +0000 (10:02 +0900)]
tools/nnapi_unittest runs as GTest (#1757)
Changes each test to GTest.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 27 Jun 2018 00:47:20 +0000 (09:47 +0900)]
Fix daily_arm_coverage build break (#1779)
Fix daily_arm_coverage build break : issue #1776
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 26 Jun 2018 23:31:19 +0000 (08:31 +0900)]
Add nnapi_unittest for ReLU (#1761)
relu_1 is for 2 ranks input model.
relu_2 is for 3 ranks input model.
relu_3 is for 4 ranks input model.
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 26 Jun 2018 10:00:42 +0000 (19:00 +0900)]
Extend broadcasting for ADD operation (#1765)
* Extend broadcating for ADD operation
This commit extends broadcating for ADD operation.
-Only if input are initialized, it supports various broadcasting.
-For example, in case left = {N=1,H=3,W=2,C=4}
these cases(right={4} or right={244} or right=3,2,4}) are supported.
-Previously, only right={1} is supported.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Remove unused blank and rename variables
This commit removes unused blank and rename variables.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Remove addShapeForBroadcast and revise addShapeConstr
This commit removes addShapeForBroadcast and revises addShapeConstr.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Change it->second.rank() to 4 (constant)
This commit changes it->second.rank() to 4
-because it always 4.
-Broadcasting tensor shape is needed only for 4D input.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Revise an incorrect comment
This commit modifies an incorrect comment.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Update and modify wrong format
This commit updates wrong format.
Signed-off-by: SungJin Choi lotieye.choi@samsung.com
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Delete duplicate addShapeConstr statement
This commit deletes duplicate addShapeConstr statement.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 26 Jun 2018 05:36:21 +0000 (14:36 +0900)]
Introduce broadcasting testcase for ADD operation (#1768)
* Introduce broadcasting cast for ADD operation
This commit itroduces more broadcasting cast for ADD operation.
-to test #1765
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* add a line at the end
This commit adds a line at the end.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 26 Jun 2018 03:53:22 +0000 (12:53 +0900)]
Add Kernel ACL test case for ReLU (#1763)
Add tests for both NEON and CL kernel were added
Support tests for 2d-tensors and 4d-tensors
KernelACL_TC.reluFloat32_2d
KernelACL_TC.reluFloat32_4d
KernelACL_TC.neon_reluFloat32_2d
KernelACL_TC.neon_reluFloat32_4d
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 26 Jun 2018 00:20:39 +0000 (09:20 +0900)]
Set parameter for ReLU in pure acl runtime (#1759)
* Set parameter for ReLU in pure acl runtime
Set parameter for ReLU in pure acl runtime compilation step
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Remove irrelevant code
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 25 Jun 2018 05:31:59 +0000 (14:31 +0900)]
Implement DEQUANTIZE operation of NNAPI in pureacl. (#1740)
This commit implements DEQUANTIZE operation in pureacl.
- Enable DEQUANTIZE operator on nnapi_delegate of tflite.
- Implement DEQUANTIZE operation in pureacl
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 25 Jun 2018 04:27:38 +0000 (13:27 +0900)]
Support general brodcasting on ADD (#1733)
* Support general brodcasting on ADD
This commit supports general brodcasting on ADD.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Modify comparison between shape
This commit modifies comparison between shape.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 25 Jun 2018 02:11:34 +0000 (11:11 +0900)]
Define ReLU namespace in pure acl runtime (#1737)
Define ReLU namespace, node, param in pure acl runtime
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 22 Jun 2018 07:16:45 +0000 (16:16 +0900)]
DIV TFLITE file Generator for testing (#1742)
This file generates the following tflite files for DIV.
These are used for #1741.
```
$ ls -1 models/div/*.tflite
models/div/div_1d_1d.tflite
models/div/div_1d_scalarConst.tflite
models/div/div_2d_1d.tflite
models/div/div_2d_2d.tflite
models/div/div_2d_scalarConst.tflite
models/div/div_3d_1d.tflite
models/div/div_3d_2d.tflite
models/div/div_3d_3d.tflite
models/div/div_3d_scalarConst.tflite
models/div/div_4d_1d.tflite
models/div/div_4d_2d.tflite
models/div/div_4d_3d.tflite
models/div/div_4d_4d.tflite
models/div/div_4d_scalarConst.tflite
models/div/div_scalarConst_scalarConst.tflite
```
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 22 Jun 2018 06:21:07 +0000 (15:21 +0900)]
Support QASYMM8 type to initializer of CONV2D op and FC op (#1710)
This commit supports QASYMM8 type to initializer of CONV2D op and FC op.
- Add initKernelTensor() function.
- Support QASYMM8 type to initializer of CONV2D op and FC op.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 22 Jun 2018 06:14:56 +0000 (15:14 +0900)]
[tool] PB info printer (#1749)
This tool prints information of operations inside a PB file.
Currently this information will be used to calculate overhead of Conv in ADAS model.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 22 Jun 2018 06:13:49 +0000 (15:13 +0900)]
Add a unit test for dequantize op. (#1735)
This commit adds a unit test for dequantize op.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 22 Jun 2018 04:12:08 +0000 (13:12 +0900)]
[Emergency] Fix build break (#1756)
Fix build break in compilation.cc
- build.append()
- scala -> scalar
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 22 Jun 2018 03:48:38 +0000 (12:48 +0900)]
Add testcases for ADD operation (#1734)
This commit adds test cases for ADD operation to check various broadcasting cases.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 22 Jun 2018 03:40:26 +0000 (12:40 +0900)]
Scale for Softmax (#1728)
Parent issue: #1675
This commit enables scale (beta) for `Softmax` operation.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 22 Jun 2018 03:40:00 +0000 (12:40 +0900)]
Introduce ReLU1 to pure acl runtime (#1715)
This commit adds ReLU1 to pure acl runtime.
-min(max(x, -1), 1);
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 22 Jun 2018 03:38:56 +0000 (12:38 +0900)]
Add activiation to avgpool and maxpool (#1718)
This commit adds activiation to avgpool and maxpool in pure acl runtime.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 22 Jun 2018 02:40:27 +0000 (11:40 +0900)]
Move functions related to make TensorInfo to Cast.h (#1705)
* Move functions related to make TensorInfo to Cast.h
This commit moves functions related to make TEnsorInfo from compilation.cc to Cast.h
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
* Add quantization information to an asTensorInfo() for operand::Shape
This commit adds quantization information to an asTensorInfo() for operand::Shape.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 22 Jun 2018 00:47:39 +0000 (09:47 +0900)]
Update ACL submodule (#1751)
Update ACL submodule: Fix error on MV8890
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 22 Jun 2018 00:47:18 +0000 (09:47 +0900)]
Fix CMakeLists for android build (#1750)
Install JNI so file to lib/
Link log for debug
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 20 Jun 2018 23:41:12 +0000 (08:41 +0900)]
Add testcases for sub operation (#1730)
This commit adds test cases for sub operation to check various broadcasting cases.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 20 Jun 2018 08:19:14 +0000 (17:19 +0900)]
Generator - TFLITE file, Tensorflow model file (#1690)
This commit enhances `tools/tensorflow_model_freezer`.
Now a lot of ranks can be used as operands and files for them are generated on the fly.
Running a sample file, 'Mul_gen.py' will creates the following files:
- __TFLITE__ files
- Tensorflow __model__ file
- __Tensorboard__ files
- __frozen__ Tensorflow model file
Please read `reamd.me` for details.
```
$ ls ~/model/mul/ -a
. mul_2d_1d.pb mul_3d_1d.pb
.. mul_2d_1d.pbtxt mul_3d_1d.pbtxt
checkoiint mul_2d_1d.tflite mul_3d_1d.tflite
mul_1d_1d_frozen.pb mul_2d_2d_frozen.pb mul_3d_3d_frozen.pb
mul_1d_1d_frozen.pbtxt mul_2d_2d_frozen.pbtxt mul_3d_3d_frozen.pbtxt
mul_1d_1d.pb mul_2d_2d.pb mul_3d_3d.pb
mul_1d_1d.pbtxt mul_2d_2d.pbtxt mul_3d_3d.pbtxt
mul_1d_1d.tflite mul_2d_2d.tflite mul_3d_3d.tflite
mul_1d_scalar_frozen.pb mul_2d_scalar_frozen.pb mul_scalar_scalar_frozen.pb
mul_1d_scalar_frozen.pbtxt mul_2d_scalar_frozen.pbtxt mul_scalar_scalar_frozen.pbtxt
mul_1d_scalar.pb mul_2d_scalar.pb mul_scalar_scalar.pb
mul_1d_scalar.pbtxt mul_2d_scalar.pbtxt mul_scalar_scalar.pbtxt
mul_1d_scalar.tflite mul_2d_scalar.tflite mul_scalar_scalar.tflite
mul_2d_1d_frozen.pb mul_3d_1d_frozen.pb .tensorboard
mul_2d_1d_frozen.pbtxt mul_3d_1d_frozen.pbtxt
```
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 20 Jun 2018 07:30:06 +0000 (16:30 +0900)]
FloatVar to handle float input of NNAPI_UNITTEST (#1720)
This commit adds FloatVar. This is needed to handle float input of NNAPI_UNITTEST for beta value of ```SOFTMAX```. (#1675)
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 20 Jun 2018 02:32:35 +0000 (11:32 +0900)]
Modify asTensorShape to support general brodcasting on SUB (#1729)
This commit modifies asTensorShape to support general brodcasting on SUB
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 20 Jun 2018 02:26:20 +0000 (11:26 +0900)]
Add README.md file of test_driver.sh (#1719)
* Add README.md file of test_driver.sh
Add the contents of how the `test_driver.sh` works.
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Update contents for test_driver
Add description for `nnapi_test` and TC directory to clarify
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 20 Jun 2018 02:22:34 +0000 (11:22 +0900)]
Update acl submodule patch (
f52df29a919ee59c4c734d3a9bd3ed6c8c0182c2) (#1731)
This commit updates acl submodule patch (
f52df29a919ee59c4c734d3a9bd3ed6c8c0182c2)
- https://github.sec.samsung.net/RS7-RuntimeNTools/ComputeLibrary/pull/27
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 20 Jun 2018 01:17:53 +0000 (10:17 +0900)]
Fix typo asScala to asScalar (#1724)
I would like to fix it because same typos spread to several places.
(Perhaps for avoiding compilation error)
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 19 Jun 2018 10:48:24 +0000 (19:48 +0900)]
Naming change: pure -> tfl_interp (#1725)
Current naming ```pure``` is confusing since we have ```pure_arm_compute```.
Since this is actually tensorflow lite interpreter, ```pure``` was renamed to ```tfl_interp```.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 19 Jun 2018 10:47:28 +0000 (19:47 +0900)]
[Pure ACL] MUL supports more operand ranks (#1687)
`MUL(A, B)` tested with the more shapes of operands.
- 1D, 2D, 3D combination
- A == 1D, 2D, 3D, B == scalar (or vice versa)
The following has issue so far (will be fixed as a separate PR):
- scalar output
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 19 Jun 2018 09:37:29 +0000 (18:37 +0900)]
Update submodule patch of ACL (#1709)
This commit updates submodule patch of ACL
- update c88554c (introduce gather 1D input)
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 19 Jun 2018 09:35:33 +0000 (18:35 +0900)]
Install nnapi_unittest_xxx files to out/unittest directory (#1723)
These files have not been run through the `test_driver.sh`.
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 19 Jun 2018 02:40:57 +0000 (11:40 +0900)]
Support other types for initializer (#1713)
This commit supports other types for initializer.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 19 Jun 2018 00:33:47 +0000 (09:33 +0900)]
Update document for docker info. (#1714)
* Update document for docker setting
Fix the wrong docker preference file location
The default location of daemon.json is `/etc/docker/daemon.json` on Linux.
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Update document for docker information
Add 'How to install Docker' section
Add HTTP/HTTPS proxy setting informations
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 18 Jun 2018 10:52:53 +0000 (19:52 +0900)]
[Pure CL] Store operation name (#1689)
This commit revises IExecutionBuilder to require the name of operations
to be added in order to faciliate debugging and profiling.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 18 Jun 2018 01:33:07 +0000 (10:33 +0900)]
Remove unused kernel files *spectrogram* and *mfcc* (#1708)
Glob in REMOVE_ITEM does not work. This commit specifies exact filenames.
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 18 Jun 2018 00:07:12 +0000 (09:07 +0900)]
[Pure CL] Extract FeatureSink (#1704)
This commit extracts FeatureSink class from execution.cc.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 09:57:32 +0000 (18:57 +0900)]
[Pure CL] Use CLArithmeticAddition by default (#1700)
This commit revises pure CL runtime to use CLArithmeticAddtiion by
default.
Now, one may enable SimpleArithmeticAddition by setting
USE_SIMPLE_ARITHMETIC_ADDITION as 1.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 09:56:04 +0000 (18:56 +0900)]
[Pure CL] Introduce Synchronization (#1693)
* [Pure CL] Introduce Synchronization
This commit introduces PURE_ARM_COMPUTE_SYNC_ENABLE option which synchronize
each operaation when it is turned on.
This option is neccessary for profiler.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Enable sync when value is non-zero
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 08:13:52 +0000 (17:13 +0900)]
Submodule update : acl and tensorflow (#1706)
1. acl :
cf4053637c8b56
2. tensorflow :
040ac56a2db2b
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
박세희/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 15 Jun 2018 07:34:27 +0000 (16:34 +0900)]
Add liblog for android build (#1702)
This will add link to liblog for android builds for tflite and pureacl
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 15 Jun 2018 06:59:44 +0000 (15:59 +0900)]
Revert "[Demo] Enable extended op for android jni (#1653)" (#1685)
This reverts commit
eded8f893006b679c4d259632523a1f108feb3b7.
Now this is not needed becasue upstream TFLite would be changed.
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 15 Jun 2018 06:24:12 +0000 (15:24 +0900)]
Modify wrong comment and typos (#1688)
This commit modifies wrong comment and typos
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 03:29:01 +0000 (12:29 +0900)]
Remove unused kernel files from TFLITE_SRCS (#1691)
1. remove spectrogram*.cc
2. remove mfcc*.cc
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 15 Jun 2018 01:56:02 +0000 (10:56 +0900)]
Add gather nnapi unittest (2D input) (#1676)
* add gather nnapi unittest (2D input)
This commit adds gather nnapi unittest.
-2D input
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Modify wrong comment and typos
This commit modify wrong comment and typos
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 15 Jun 2018 01:25:26 +0000 (10:25 +0900)]
Add assert statement for gather. (#1683)
This commit adds assert statement for gather.
-Gather supports 1D-input and 2D input
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 01:25:08 +0000 (10:25 +0900)]
[Pure CL] Rename as SimpleArithmeticAddition (#1686)
This commit renames SimpleArithmeticAdditionLayer as
SimpleArithmeticAddition to align with the corresponding CL
implementation (CLArithmeticAddition).
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 15 Jun 2018 01:24:16 +0000 (10:24 +0900)]
Modify initialization for gather operation (#1666)
This commit modifies initialization for gather operation.
- initialize from 0, not 1.
- because gather indices' values should be within input data coverage.
- for example, suppose input data[300], 0=< indices data's value <299.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
서상민/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 00:56:06 +0000 (09:56 +0900)]
Add a unit test for int32-to-float32 conversion (#1643)
This patch adds a unit test that tests a cast operation from int32
tensor to float32 tensor.
Signed-off-by: Sangmin Seo <sangmin7.seo@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 00:37:32 +0000 (09:37 +0900)]
[Pure CL] asTensor3DSink (#1684)
This commit extracts rank-3 specialization in asTensorSink as
asTensor3DSink method, and call it from setOutput API.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 00:33:16 +0000 (09:33 +0900)]
[Pure CL] Introduce Step class (#1670)
This commit introduces Step class which wraps arm_compute::IFunction.
Step class can record some meaningful name along with
arm_compute::IFunction object which may facilitate debugging and
performance analysis.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 00:29:17 +0000 (09:29 +0900)]
[Pure CL] Add 'FeatureLoggingLayer' (#1672)
This commit adds 'FeatureLoggingLayer' which prints the content of
'ICLTensor' as a feature.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 00:22:47 +0000 (09:22 +0900)]
[Pure CL] Extract 'VectorSink' class (#1671)
This commit extracts 'VectorSink' class into a separate file.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 15 Jun 2018 00:17:31 +0000 (09:17 +0900)]
ACL submodule update (#1673)
ACL submodule update to latest version
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 14 Jun 2018 11:10:27 +0000 (20:10 +0900)]
Support other types of operand for SimpleArithmeticAdditionLayer (#1668)
This commit supports other types of operand for SimpleArithmeticAdditionLayer.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 14 Jun 2018 11:09:58 +0000 (20:09 +0900)]
Optimized TensorSink (#1665)
Parent issue : #1658
memcpy is used just like #1647.
Tested with `tools/nnapi_unittests/test/mul_1` and `mul_2`.
Signed-off-by: Hyun Sik Yoon hyunsik.yoon@samsung.com
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 14 Jun 2018 11:08:43 +0000 (20:08 +0900)]
[tflite_benchmark] Show min/max (#1680)
This commit revises tflite_benchmakr to show min/max when benchmark
finishes.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 14 Jun 2018 11:08:35 +0000 (20:08 +0900)]
[tflite_benchmark] Show Input/Output names (#1681)
This commit revises tflite_benchmark to show input/output tensor name to
make it easy to understand the model under analysis.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 14 Jun 2018 11:05:45 +0000 (20:05 +0900)]
Environment variable to use upstream delegate (#1682)
* Environment variable to use upstream delegate
Define environment variable to use upstream delegate
Need to set UPSTREAM_DELEGATE=1 to use upstream delegate
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Set UseNNAPI
Call UseNNAPI(true) to set nnapi delegate flag in interpreter
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 14 Jun 2018 08:22:48 +0000 (17:22 +0900)]
[Pure ACL] added Rank 2 input (#1678)
This commit intrudocues rank 2 input.
Naming (`matrix`) was used to be aligned with `MatrixSink`
Tested with tflite file with `Mul` (2d x scalar) and `Gather`.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 14 Jun 2018 07:40:26 +0000 (16:40 +0900)]
Change quantization of test codes to default value. (#1655)
In case of the scalars and non-fixed point tensors, scale value is changed to 0.0f by TFLite because it is not used.
However in TFLITE, scale value of TENSOR_INT32 is passed on to the runtime and may be used depending on each operation.
So This commit set scale value to default value(0.0f) in order to clarify the value is not used.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 14 Jun 2018 05:17:35 +0000 (14:17 +0900)]
Introduce 'nnfw::make_unique' (#1677)
This commit deprecates 'make_layer' function in pure CL runtime, and
introduces a general make_unique function as a helper.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Thu, 14 Jun 2018 04:38:28 +0000 (13:38 +0900)]
[nnapi_unittest] more cases for MUL (3D, 1D) (#1669)
For PR #1665,
`Mul` without ` 1 ` from shape ([5, 3, 12] * [12]) was written.
Previous version `Mul_1' (for ADAS, [300, 1, 4] * [4]) sometimes worked with buggy intermediate code while working on #1665.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 14 Jun 2018 02:04:28 +0000 (11:04 +0900)]
Model parser: print tensor data type (#1674)
- Print tensor data type such as FLOAT32, INT32
- Fix readme
- Change format to show tensor name
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 12 Jun 2018 07:55:35 +0000 (16:55 +0900)]
add gather nnapi unittest (#1664)
This commit adds gather nnapi unittest.
Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 12 Jun 2018 07:54:57 +0000 (16:54 +0900)]
[Pure CL] Support ReLU6 activation (#1663)
This commit allows pure CL runtime to process ReLU6 activation (which is
used in MobileNet example).
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 12 Jun 2018 07:54:27 +0000 (16:54 +0900)]
[NNAPI Unittest] Add 'add_3' test (#1628)
This commit adds 'add_3' NNAPI unittest which allows users freely to
specify LHS/RHS shape.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 12 Jun 2018 07:53:43 +0000 (16:53 +0900)]
[Pure CL] Support DepthwiseConv2D (#1661)
This commit introduces DpethwiseConv2D operation support in pure CL
runtime.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 12 Jun 2018 07:52:46 +0000 (16:52 +0900)]
Remove setting operand type of execution input. (#1656)
This commit removes setting operand type of execution input.
It is unnecessary to reset operand type
because operand type is already set in model step
and scale of TENSOR_INT32 may be used depending on each operation.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 12 Jun 2018 02:21:13 +0000 (11:21 +0900)]
unittest for MUL (3Dx1D) (#1645)
Unittest for MUL ([300,1,4], [4]) has added.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 11 Jun 2018 23:52:32 +0000 (08:52 +0900)]
[Demo] Enable extended op for android jni (#1653)
This will enable our extended operators for TFLite JNI
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 11:56:35 +0000 (20:56 +0900)]
[Pure CL] Add 'MatrixSink' (#1647)
* [Pure CL] Add 'MatrixSink'
This commit adds 'MatrixSink' which efficiently copies output of matrix
shape (rank-2).
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
서상민/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 09:52:18 +0000 (18:52 +0900)]
[PureCL] Use ACL Cast implementation (#1651)
* [PureCL] Use ACL Cast implementation
For issue #1431
This patch updates the ACL submodule and revises the Pure CL runtime to
use ACL's CAST implementation by default.
Signed-off-by: Sangmin Seo <sangmin7.seo@samsung.com>
* Add cast to pureacl_frameworktest_list.txt
* Clean up the duplicated codes