platform/core/ml/nnfw.git
6 years agotools/nnapi_unittest runs as GTest (#1757)
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>
6 years agoFix daily_arm_coverage build break (#1779)
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>
6 years agoAdd nnapi_unittest for ReLU (#1761)
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>
6 years agoExtend broadcasting for ADD operation (#1765)
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>
6 years ago Introduce broadcasting testcase for ADD operation (#1768)
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>
6 years agoAdd Kernel ACL test case for ReLU (#1763)
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>
6 years ago Set parameter for ReLU in pure acl runtime (#1759)
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>
6 years agoImplement DEQUANTIZE operation of NNAPI in pureacl. (#1740)
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>
6 years agoSupport general brodcasting on ADD (#1733)
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>
6 years agoDefine ReLU namespace in pure acl runtime (#1737)
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>
6 years agoDIV TFLITE file Generator for testing (#1742)
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>
6 years agoSupport QASYMM8 type to initializer of CONV2D op and FC op (#1710)
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>
6 years ago[tool] PB info printer (#1749)
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>
6 years agoAdd a unit test for dequantize op. (#1735)
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>
6 years ago[Emergency] Fix build break (#1756)
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>
6 years agoAdd testcases for ADD operation (#1734)
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>
6 years agoScale for Softmax (#1728)
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>
6 years agoIntroduce ReLU1 to pure acl runtime (#1715)
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>
6 years agoAdd activiation to avgpool and maxpool (#1718)
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>
6 years agoMove functions related to make TensorInfo to Cast.h (#1705)
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>
6 years agoUpdate ACL submodule (#1751)
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>
6 years agoFix CMakeLists for android build (#1750)
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>
6 years agoAdd testcases for sub operation (#1730)
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>
6 years agoGenerator - TFLITE file, Tensorflow model file (#1690)
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>
6 years agoFloatVar to handle float input of NNAPI_UNITTEST (#1720)
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>
6 years agoModify asTensorShape to support general brodcasting on SUB (#1729)
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>
6 years agoAdd README.md file of test_driver.sh (#1719)
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>
6 years agoUpdate acl submodule patch (f52df29a919ee59c4c734d3a9bd3ed6c8c0182c2) (#1731)
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>
6 years agoFix typo asScala to asScalar (#1724)
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>
6 years agoNaming change: pure -> tfl_interp (#1725)
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>
6 years ago[Pure ACL] MUL supports more operand ranks (#1687)
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

6 years agoUpdate submodule patch of ACL (#1709)
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>
6 years agoInstall nnapi_unittest_xxx files to out/unittest directory (#1723)
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>
6 years agoSupport other types for initializer (#1713)
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>
6 years agoUpdate document for docker info. (#1714)
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>
6 years ago[Pure CL] Store operation name (#1689)
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>
6 years agoRemove unused kernel files *spectrogram* and *mfcc* (#1708)
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>
6 years ago[Pure CL] Extract FeatureSink (#1704)
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>
6 years ago[Pure CL] Use CLArithmeticAddition by default (#1700)
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>
6 years ago[Pure CL] Introduce Synchronization (#1693)
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

6 years agoSubmodule update : acl and tensorflow (#1706)
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>
6 years agoAdd liblog for android build (#1702)
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>
6 years agoRevert "[Demo] Enable extended op for android jni (#1653)" (#1685)
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>
6 years agoModify wrong comment and typos (#1688)
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>
6 years agoRemove unused kernel files from TFLITE_SRCS (#1691)
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>
6 years agoAdd gather nnapi unittest (2D input) (#1676)
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

6 years agoAdd assert statement for gather. (#1683)
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>
6 years ago[Pure CL] Rename as SimpleArithmeticAddition (#1686)
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>
6 years agoModify initialization for gather operation (#1666)
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>
6 years agoAdd a unit test for int32-to-float32 conversion (#1643)
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>
6 years ago[Pure CL] asTensor3DSink (#1684)
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>
6 years ago[Pure CL] Introduce Step class (#1670)
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>
6 years ago[Pure CL] Add 'FeatureLoggingLayer' (#1672)
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>
6 years ago[Pure CL] Extract 'VectorSink' class (#1671)
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>
6 years agoACL submodule update (#1673)
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>
6 years agoSupport other types of operand for SimpleArithmeticAdditionLayer (#1668)
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>
6 years agoOptimized TensorSink (#1665)
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
6 years ago[tflite_benchmark] Show min/max (#1680)
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>
6 years ago[tflite_benchmark] Show Input/Output names (#1681)
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>
6 years agoEnvironment variable to use upstream delegate (#1682)
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>
6 years ago[Pure ACL] added Rank 2 input (#1678)
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>
6 years agoChange quantization of test codes to default value. (#1655)
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>
6 years agoIntroduce 'nnfw::make_unique' (#1677)
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>
6 years ago[nnapi_unittest] more cases for MUL (3D, 1D) (#1669)
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>
6 years agoModel parser: print tensor data type (#1674)
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>
6 years agoadd gather nnapi unittest (#1664)
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>
6 years ago[Pure CL] Support ReLU6 activation (#1663)
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>
6 years ago[NNAPI Unittest] Add 'add_3' test (#1628)
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>
6 years ago[Pure CL] Support DepthwiseConv2D (#1661)
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>
6 years agoRemove setting operand type of execution input. (#1656)
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>
6 years agounittest for MUL (3Dx1D) (#1645)
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>
6 years ago[Demo] Enable extended op for android jni (#1653)
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>
6 years ago[Pure CL] Add 'MatrixSink' (#1647)
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>
6 years ago[PureCL] Use ACL Cast implementation (#1651)
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

6 years agoEnable CAST framework test (#1649)
서상민/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 09:19:21 +0000 (18:19 +0900)]
Enable CAST framework test (#1649)

This patch enables the CAST framework test as the CPU fallback and the
Pure CL runtime can execute it correctly.

Signed-off-by: Sangmin Seo <sangmin7.seo@samsung.com>
6 years agoImplement CAST operation for TENSOR_QUANT8_ASYMM in CPU fallback (#1609)
서상민/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 07:48:01 +0000 (16:48 +0900)]
Implement CAST operation for TENSOR_QUANT8_ASYMM in CPU fallback (#1609)

For issue #1431 and #1199

This patch implements CAST operation for TENSOR_QUANT8_ASYMM in CPU
fallback.  This implementation will be used for comparison with that of
the Pure CL runtime.

Signed-off-by: Sangmin Seo <sangmin7.seo@samsung.com>
6 years agoAdd quantization information to TensorInfo of CAST (#1646)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 11 Jun 2018 07:24:09 +0000 (16:24 +0900)]
Add quantization information to TensorInfo of CAST (#1646)

This commit adds quantization information to TensorInfo of CAST.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
6 years agoInitialize input values in tflite_benchmark (#1631)
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 11 Jun 2018 07:19:34 +0000 (16:19 +0900)]
Initialize input values in tflite_benchmark (#1631)

This commit initializes input values in tflite_benchmark file.
-because gather operation fails without inititalization

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
6 years ago[PureACL] Add QuantizationInfo to TensorInfo (#1638)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 11 Jun 2018 06:31:57 +0000 (15:31 +0900)]
[PureACL] Add QuantizationInfo to TensorInfo (#1638)

This commit adds QuantizationInfo to TensorInfo for supporting to quantize.
Added QuantizationInfo enables to use qauntize(float to quantized uint8) and deqauntize(quantized uint8 to float) when implementing each Op using the arm_compute library.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
6 years agoPure CL: initialize 2D operand workaround (#1641)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 06:21:57 +0000 (15:21 +0900)]
Pure CL: initialize 2D operand workaround (#1641)

Initialize 2D operand
Workarund using vector initialize

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
6 years ago[PureACL] Enable QASYMM8 type for Cast operation (#1644)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 11 Jun 2018 06:21:14 +0000 (15:21 +0900)]
[PureACL] Enable QASYMM8 type for Cast operation (#1644)

This commit enables QASYMM8 type for Cast operation in CPU only.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
6 years ago[PureACL] Support other type of operand for setting tensorinput in execution (#1642)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 11 Jun 2018 06:19:57 +0000 (15:19 +0900)]
[PureACL] Support other type of operand for setting tensorinput in execution (#1642)

This commit supports other type of operand for setting tensorinput in execution.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
6 years ago[NNAPI unittest] Add 'dconv_1' testcase (#1601)
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 05:37:59 +0000 (14:37 +0900)]
[NNAPI unittest] Add 'dconv_1' testcase (#1601)

* [NNAPI unittest] Add 'dconv_1' testcase

This commit adds 'dconv_1' testcase which allows users to test various
depthwise-convolution examples.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Do NOT update off inside loop

6 years ago[Pure CL] Use squeezed shape in setOutput (#1630)
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 05:31:45 +0000 (14:31 +0900)]
[Pure CL] Use squeezed shape in setOutput (#1630)

* [Pure CL] Use squeezed shape in setOutput

This commit simplifies ANeuralNetworksExecution_setOutput via using
squeezed shape.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Update comments

6 years agoImplement the plan of Cast operation (#1574)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 11 Jun 2018 05:28:15 +0000 (14:28 +0900)]
Implement the plan of Cast operation (#1574)

This commit implements the plan of Cast operation.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
6 years agoTFLiteSharp code for running model inference with test App (#1486)
Kshitiz Bansal/Advanced SW Team /SRI-Delhi/Engineer/삼성전자 [Mon, 11 Jun 2018 05:24:00 +0000 (10:54 +0530)]
TFLiteSharp code for running model inference with test App (#1486)

This patch contains code for the TFLiteSharp's Run function. Currently it supports only one dimensional inputs of the types INT32, INT64, FLOAT32 and UINT8.
A new VS project for testing the TFLiteSharp wrapper functions is also added. The project is a xamarin app that creates an interpreter instance using preloaded mobilenet model and
runs the inference using a bitmap image as an input.

Signed-off-by: Kshitiz Bansal <kshitiz.b@samsung.com>
6 years agoEnable a framework test for STRIDED_SLICE (#1637)
최형규/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 04:29:39 +0000 (13:29 +0900)]
Enable a framework test for STRIDED_SLICE (#1637)

- Enable a framework test for STRIDED_SLICE

Signed-off-by: Hyung-Kyu Choi <hk0110.choi@samsung.com>
6 years agoPure CL: 2D Concat (#1636)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 02:52:31 +0000 (11:52 +0900)]
Pure CL: 2D Concat (#1636)

Support 2D concat (axis: row)

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
6 years agoSubmodule update (acl max, stridedlisce) + acl test added (#1635)
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 11 Jun 2018 02:32:58 +0000 (11:32 +0900)]
Submodule update (acl max, stridedlisce) + acl test added (#1635)

1. Submodule update to 39cc3ac5371d
2. Add tesnsorflowmax, mul/broadcast, strided_slice into pureacl test list

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
6 years agoAdd a benchmark per operation to test_driver.sh (#1589)
최성진/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 11 Jun 2018 02:09:51 +0000 (11:09 +0900)]
Add a benchmark per operation to test_driver.sh (#1589)

* introduce benchmark per operation with tflite model

This commit add new benchmark per operation to test_driver with tflite model.

 -$ ./tools/test_driver/test_driver.sh --artifactpath=. --benchmark_op
 -$ ./tools/test_driver/test_driver.sh --artifactpath=. --benchmark_op ----frameworktest_list_file="specific file path that contains model lists(i.e., add/4D softmax gather topkv2 etc.)"

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Modify comment to show operation test

This commit modifies comment to show operation test

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Generate benchmark_op_result.json file in test_driver.sh

This commit generates benchmark_op_result.json file in test_driver.sh
-benchmark_op_result.json (for benchmark per operation) as well as benchmark_result.json

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* fix setting pure acl lib path

- remove setting LD_LIBRARY_PATH in run_benchmark_op.sh
- pass pure acl lib path to run_test.sh in run_benchmark_op.sh
- modify setting NNFW_LD_LIBRARY_PATH in run_test.sh

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* remove unused parameter OPERATION_TEST

This commit removes unused parameter OPERATION_TEST anymore.

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* Modify list file of benchmark-operation

This commit modifies list file of benchmark-operation
-The framework test uses pureacl_frameworktest_list.txt
-The benchmark operation test uses benchmark_op_list.txt
-The benchmark_op_list is different from benchmark_op_list.txt because gather is not yet applied to include/kerenl/acl/Gather.h (#1535 due to tizen bulid failure). In the end, they will be same.

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* modify USE_NNAPI env parameter.

This commit modifies USE_NNAPI environment parameter.
- TFLite+NNRuntime+ACL-OpenCL value was wrong becuase of USE_NNAPI

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* remove download and verify part

This commit remove downlaod and verify parts because run_test.sh has these parts.

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
* remove unnecessary blank and underline

This commit removes unnecessary blank and underline

Signed-off-by: SungJin Choi <lotieye.choi@samsung.com>
6 years ago[Pure CL] Remove tensor type conversion (#1629)
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Sun, 10 Jun 2018 23:37:46 +0000 (08:37 +0900)]
[Pure CL] Remove tensor type conversion (#1629)

This conversion is no nolonger necessary as operand::Shape now inherits tensor::Shape.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
6 years agoFix BUILD break on push #304 (#1633)
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Sun, 10 Jun 2018 23:36:58 +0000 (08:36 +0900)]
Fix BUILD break on push #304 (#1633)

Fix : [CI] push-builder #304 failed!

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
6 years agoAdd 'prepend' method to 'TensorShape' (#1620)
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Sun, 10 Jun 2018 23:36:12 +0000 (08:36 +0900)]
Add 'prepend' method to 'TensorShape' (#1620)

This commit introduces 'preprend' method to 'TensorShape' to make it
easy to handle broadcasting in Tensorflow Lite.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
6 years agoIntroduce RandomTestRunner factory method (#1618)
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Sun, 10 Jun 2018 23:35:49 +0000 (08:35 +0900)]
Introduce RandomTestRunner factory method (#1618)

* Introduce RandomTestRunner factory method

This commit introduces RandomTestRunner::make method which construct
RandomTestRunner with parameters specified by VERBOSE and TOLERANCE
environment variables.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
6 years agoMUL(3Dx1D) for PureACL (#1570)
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Fri, 8 Jun 2018 10:03:43 +0000 (19:03 +0900)]
MUL(3Dx1D) for PureACL (#1570)

* MUL(3Dx1D) for PureACL

Parent issue #1374

This is MUL(3D, 1D) operation for ADAS model.
This needs to be enhanced after writing generic tensor support.
Currently this is written by piggybacking code of `Feature`.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* MUL(3Dx1D) for PureACL

Parent issue #1374

This is MUL(3D, 1D) operation for ADAS model.
This also includes generic tensor support for MUL.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
6 years ago[Pure CL] Remove unused shape field/method (#1625)
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 8 Jun 2018 08:14:03 +0000 (17:14 +0900)]
[Pure CL] Remove unused shape field/method (#1625)

This commit removed unused (and unnecessary) shape fields and method
from ARMCompute view implementations.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
6 years agoSet as_dims as static inline function (#1621)
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 8 Jun 2018 08:07:06 +0000 (17:07 +0900)]
Set as_dims as static inline function (#1621)

This commit adds 'static inline' quantifier to as_dims method to avoid
multiple definition problems.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
6 years ago[PureCL] Support TENSOR_QUANT8_ASYMM type operand (#1623)
서상민/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 8 Jun 2018 08:06:56 +0000 (17:06 +0900)]
[PureCL] Support TENSOR_QUANT8_ASYMM type operand (#1623)

For issue #1481

This patch revises `internal::tflite::operand::Object` to keep `scale`
and `zeroPoint` in order to support operand of
ANEURALNETWORKS_TENSOR_QUANT8_ASYMM type, and then modifies
`ANeuralNetworksModel_addOperand` to pass the scale and zeroPoint values
as well when appending a new operand.

Signed-off-by: Sangmin Seo <sangmin7.seo@samsung.com>
6 years agoUse 'static_nnfw_util' (#1624)
박종현/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 8 Jun 2018 08:06:35 +0000 (17:06 +0900)]
Use 'static_nnfw_util' (#1624)

This commit revises pure CL runtime to use static nnfw_util library
instead of shared one (to make it easy to integrate into android app).

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
6 years ago[PureCL] Handle TENSOR_QUANT8_ASYMM conversion (#1622)
서상민/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 8 Jun 2018 07:02:21 +0000 (16:02 +0900)]
[PureCL] Handle TENSOR_QUANT8_ASYMM conversion (#1622)

For issue #1481

This patch revises conversion functions to handle
ANEURALNETWORKS_TENSOR_QUANT8_ASYMM as well.

Signed-off-by: Sangmin Seo <sangmin7.seo@samsung.com>