박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 07:49:29 +0000 (16:49 +0900)]
[locop] Use const loco::Node * (if possible) (#3946)
This commit rewrites some code that takes loco::Node * as input to take
const loco::Node * instead.
This is a preliminary step to use loco visitor.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 07:44:42 +0000 (16:44 +0900)]
[loco] Simplify CanonicalNodeVisitor default fallback (#3945)
This commit simplifies default visitor fallback implementation.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 24 Jun 2019 07:36:51 +0000 (16:36 +0900)]
[locomotiv] constant input for avgpool2d test (#3934)
* [locomotiv] constant input for avgpool2d test
This will revise avgpool2d test with constant input for the python example code in the comment for verification.
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* make runnable
* remove unnecessary and match indentation
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 24 Jun 2019 07:32:00 +0000 (16:32 +0900)]
[moco/tf] remove unnecessary comments (#3947)
This will remove unnecessary comments in FixShapeTransfrom test.
- graph is local and cleaning of annotation will be done when graph destructor is called
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 07:29:08 +0000 (16:29 +0900)]
[moco/tf] Update Conversion.md with Identity (#3938)
* [moco/tf] Update Conversion.md with Identity
This commit updates Conversion.md to explain how to convert TensorFlow
Identity node into loco.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Use print(...)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 07:20:09 +0000 (16:20 +0900)]
[tf2tflite] Add copyright notice (#3939)
Copyright notice is currently missing in Driver.cpp.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 24 Jun 2019 07:16:17 +0000 (16:16 +0900)]
[moco/tf] revisit graph cleanup (#3940)
* [moco/tf] revisit graph cleanup
This will remove prepare and cleanup transformation in graph load method
- shape inference annotation can be reused after load with this
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* rollback mem leak comments
* undo remove prepare
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 06:21:08 +0000 (15:21 +0900)]
Extract enco-intf from enco (#3936)
This commit extracts enco-intf from enco as the first step
toward enco directory layout revision.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 24 Jun 2019 06:08:54 +0000 (15:08 +0900)]
[moco/tf] Enable mocotest-tf tests (#3941)
This commit will enable mocotest-tf tests,
AvgPool_000, AvgPool_001, Conv2D_000
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 05:56:34 +0000 (14:56 +0900)]
[moco/tf] Introduce configuration infrastructure (#3935)
This commit introduces basic configuration infrastructure which allows
users to control moco's internal behavior through environment variables.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 24 Jun 2019 04:57:11 +0000 (13:57 +0900)]
[locomotiv] Conv2D support padding (#3814)
* [locomotiv]Conv2D support padding
This commit adds a test for padding.
Signed-off-by: seongwoo <sw4670.chae@samsung.com>
* This commit add some example for padding
* This commit modify some indentation.
* This commit moidfy indentation.
* This commit is adding comment in test script to help understand.
* This commit add tensorflow code.
* This commit modify comment for test
* This commit apply comments.
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 24 Jun 2019 04:37:57 +0000 (13:37 +0900)]
[moco/tf] introduce graph cleanup (#3932)
* [moco/tf] introduce graph release
This will introduce release method for loco::Graph that will cleanup memories that was allocated in load method such as node annotations.
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use name cleanup
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 04:25:34 +0000 (13:25 +0900)]
Add check-copyright command (#3927)
This commit adds check-copyright command which checkes whether the
copyrigh notice is present or not.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 24 Jun 2019 03:28:10 +0000 (12:28 +0900)]
[moco/tf] Introduce BiasAdd tests for moco/tf (#3922)
This commit will introduce BiasAdd_000 and BiasAdd_001 tests for `moco/tf`
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 03:27:57 +0000 (12:27 +0900)]
[angkor] Use GTest_AddTest (#3921)
Now, angkor's build script uses GTest_AddTest without referring
gtest_main.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 24 Jun 2019 03:27:46 +0000 (12:27 +0900)]
[moco] Remove empty test directory (#3925)
Now all the tests are in mocotest projects.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 24 Jun 2019 03:14:06 +0000 (12:14 +0900)]
[locomotiv] enable AvgPool2D (#3929)
This will enable AvgPool2D node in locomotiv
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 24 Jun 2019 01:20:58 +0000 (10:20 +0900)]
[moco/tf] Introduce NodeShape (#3926)
* [moco/tf] Introduce NodeShape
This will introduce NodeShape to access shape inference result(shape information) of each loco::Node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* apply comments
* apply another comment
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 24 Jun 2019 00:07:11 +0000 (09:07 +0900)]
[moco/tf] revise test for shape inf feature (#3931)
This will revise feature_shape accessors test in shape inference data to remove implementation detail and focus on interface testing
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Sun, 23 Jun 2019 23:56:30 +0000 (08:56 +0900)]
[pepper-strcast] Use proper target name (#3917)
pepper-strcast currently exposes itself as pepper-string.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Sun, 23 Jun 2019 23:55:18 +0000 (08:55 +0900)]
Extract moco-onnx from moco (#3919)
This commit extracts moco-onnx from moco in order to make it easy to
manage internal dependencies.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 23 Jun 2019 23:48:20 +0000 (08:48 +0900)]
[moco/tf] revise test for shape inf bias (#3930)
This will revise bias_shape accessors test in shape inference data
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 23 Jun 2019 23:46:42 +0000 (08:46 +0900)]
[mocotest/tf] add avgpool2d_001 (#3928)
This will add AvgPool_001 test for VALID padding
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 21 Jun 2019 09:36:19 +0000 (18:36 +0900)]
[moco/tf] revise test for shape inf filter (#3920)
This will revise filter_shape accessors test in shape inference data
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 08:42:36 +0000 (17:42 +0900)]
[ann-ref] Check required targets (#3923)
Now, ann-ref build gets stuck with error message if required targets are
missing.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 21 Jun 2019 07:51:46 +0000 (16:51 +0900)]
[locomotiv] add method for condition check (#3912)
There are many validation check if (...) throw ... . This commit adds a function to do this with one call.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 21 Jun 2019 07:41:48 +0000 (16:41 +0900)]
[moco/tf] support bias in ShapeInferenceData (#3913)
This will introduce bias attribute accessor in ShapeInferenceData
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 06:49:30 +0000 (15:49 +0900)]
[moco-log] Use Builtin Hermes Console Reporter (#3914)
This commit replaces locally-implemented reporter with the one in
hermes-std module.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 05:43:51 +0000 (14:43 +0900)]
Extract moco-log from moco (#3897)
Let's extract "moco-log" and manage its internal dependencies.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 05:32:02 +0000 (14:32 +0900)]
[tf2tflite] Use TargetRequire_Return (#3910)
This commit replaces ensure_required_targets (which is local to
tf2tflite) with TargetRequire_Return.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 21 Jun 2019 05:13:24 +0000 (14:13 +0900)]
[moco/tf] introduce domain in ShapeInferenceData (#3901)
* [moco/tf] introduce domain in ShapeInferenceData
This will introduce domain property in ShapeInferenceData
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* add todo comment
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 04:57:08 +0000 (13:57 +0900)]
Require moco from nnkit (#3909)
nnkit-moco-tf in nnkit needs moco, and thus this dependency should be
explicit in nnkit.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 21 Jun 2019 04:40:21 +0000 (13:40 +0900)]
[locomotiv] BiasEncode (#3888)
* [locomotiv] BiasEncode
This commit adds BiasEncode and its test cases for locomotiv.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* checking input rank == 1
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 21 Jun 2019 04:40:05 +0000 (13:40 +0900)]
[moco/tf] Setting shape of Pull from user-provided input data (#3889)
This enables shape of Pull to be filled from user-provided input data.
(Assuming the user-provided input data is already stored in CompilerArgs class)
Note that this was originally writte while Conv2D was written to figure out unknown dim from user-provided shape info.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 04:22:17 +0000 (13:22 +0900)]
Extract moco tests from moco (#3907)
This commit extracts mocotest projects (mocotest-tf and moctest-onnx)
from moco in order to break dependency cycle between moco and nnkit.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 04:19:23 +0000 (13:19 +0900)]
[onnxkit] Update README.md (#3900)
Now, onnxkit's README.md is compatible with Doxygen.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 04:07:26 +0000 (13:07 +0900)]
Add TargetRequire module (#3895)
This commit adds TargetRequire module which includes the following
CMake helpers which simplifies required target check:
- TargetRequire_Check
- TargetRequire_Assert
- TargetRequire_Return
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 03:05:22 +0000 (12:05 +0900)]
[moco.tf] Remove load_model unittest (#3905)
This commit removes "load_model" test; Now tf2tflite is responsoble for
tests of this kind.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 01:55:21 +0000 (10:55 +0900)]
[moco.tf] Add Conversion.md (#3894)
This commit adds Conversion.md which describes how to express each
T/F operation on top of loco.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 01:44:53 +0000 (10:44 +0900)]
[pepper-strcast] Initial commit (#3896)
This commit introduces pepper-strcast which will hold various
string-to-value casting helpers.
The initial version supports string-to-int casting.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 01:33:24 +0000 (10:33 +0900)]
[tfldump] Update README.md (#3899)
Now, tfldump's README.md is compatible with Doxygen.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 21 Jun 2019 00:00:09 +0000 (09:00 +0900)]
[hermes-std] Initial commit (#3893)
This commit introduces hermes-std which includes basic hermes extensions
such as console reporter.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 20 Jun 2019 23:18:53 +0000 (08:18 +0900)]
[locomotiv] BiasAdd (#3877)
* [locomotiv] BiasAdd
This commit adds BiasAdd and test case for locomotiv.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* checking loco::Domain::Tensor
* msg
* setting the domain of input[1] to loco::Domain::Bias
* fix message
* more check
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 20 Jun 2019 22:27:31 +0000 (07:27 +0900)]
[loco exporter] Naming multiple graph I/O from loco (#3880)
* [loco exporter] Naming multiple graph I/O from loco
Previously loco exporter only support single graph input and single
graph output. With this commit, it now can support multiple inputs and
outputs with I/O tensor name copied from loco.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Review feedback: better comment look
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 20 Jun 2019 10:20:11 +0000 (19:20 +0900)]
[locop] Introduce NodeDesc class (#3891)
* [locop] Introduce NodeDesc class
This commit introduces NodeDesc class and refactors the internal implementation of
LinearV1 formatter using this class.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Update per feedback
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 20 Jun 2019 09:09:40 +0000 (18:09 +0900)]
[tftestinfo] Remove T/F dependencies (#3892)
tftestinfo now does not depend on TensorFlow.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 20 Jun 2019 06:13:43 +0000 (15:13 +0900)]
[stdex] Update README.md (#3886)
This commit updates stdex's README.md to be compatible with Doxygen.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 20 Jun 2019 01:39:05 +0000 (10:39 +0900)]
[tfkit] Support conversion from pb (#3883)
tfkit convert command now supports pb-to-JSON conversion.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 20 Jun 2019 01:38:57 +0000 (10:38 +0900)]
[enco.caffe] Remove local tests (#3884)
enco caffe test framework no longer uses these local tests.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 19 Jun 2019 22:54:20 +0000 (07:54 +0900)]
[moco/tf] arrange concat validate codes (#3881)
This will move(arrange) codes in build to validate for ConcatV2
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 19 Jun 2019 17:49:10 +0000 (02:49 +0900)]
[mirunner] Update README.md (#3857)
This commit updates mirunner's README.md to be compatible with Doxygen.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 19 Jun 2019 10:55:41 +0000 (19:55 +0900)]
[enco.caffe] Use shared test suite (#3882)
This commit updates enco caffe test frameworks to use the shared
test suite introduced recently.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 19 Jun 2019 07:11:56 +0000 (16:11 +0900)]
[tf2tflite] Add BiasAdd 000 testcase (#3876)
This commit adds BiasAdd 000 testcase which includes a simple T/F network
that performs BiasAdd.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 19 Jun 2019 05:24:42 +0000 (14:24 +0900)]
[loco exporter] Support BiasAdd<Tensor> (#3875)
This commit supports BiasAdd loco node of Tensor domain for loco
exporter.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 19 Jun 2019 05:22:45 +0000 (14:22 +0900)]
[moco/tf] fix style for filter_shape (#3866)
* [moco/tf] fix style for filter_shape
This will fix style of filter_shape() like existing feature_shape()
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* add comment
* throw
* fix typo
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 19 Jun 2019 04:43:11 +0000 (13:43 +0900)]
[cli] Update README.md (#3867)
This commit updates cli's README.md to be compatible with Doxygen.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 19 Jun 2019 04:40:38 +0000 (13:40 +0900)]
[loco exporter] Support BiasEncode (#3872)
This commit supports BiasEncode loco node for loco exporter
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 19 Jun 2019 04:40:13 +0000 (13:40 +0900)]
[moco/tf] throw instead of assert (#3874)
This will use throw runtime_error if rank mismatches for ShareInferenceData
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 19 Jun 2019 03:53:45 +0000 (12:53 +0900)]
[moco/tf] FixPaddingTransform for Conv2D (#3863)
* [moco/tf] FixPaddingTransform for Conv2D
This commit adds FixPaddingTransform implementation and test cases for Conv2D.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* comment, win_-> ker
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 19 Jun 2019 03:45:58 +0000 (12:45 +0900)]
Add BVLCCaffeTest resource (#3868)
This commit adds BVLCCaffeTest resource which includes various caffe
prototxt for testing.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 11:49:54 +0000 (20:49 +0900)]
[moco/tf] FixShapeTransform for Conv2D (#3858)
This commit adds FixShapeTransform for Conv2D and two test cases extracted from Inception_v3.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 18 Jun 2019 11:49:16 +0000 (20:49 +0900)]
[loco exporter] Support TensorConcat (#3861)
This commit supports TensorConcat loco node for loco exporter.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 11:42:12 +0000 (20:42 +0900)]
[tf2tflite] Add TF_IV3_Conv2D_000 (#3859)
This commit adds TF_IV3_Conv2D_000 test which simulates
"InceptionV3/InceptionV3/Conv2d_1a_3x3/Conv2D" node in inception v3
model.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 18 Jun 2019 10:44:13 +0000 (19:44 +0900)]
[loco exporter] Support Conv2D (#3849)
This commit supports Conv2D node for loco exporter
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 10:11:40 +0000 (19:11 +0900)]
[moco/tf] Methods to handle feature in ShapeInferenceData (#3855)
* [moco/tf] Methods to handle feature in ShapeInferenceData
This commit adds getter and setter for feature into/from ShapeInferenceData.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* HWCN to NHWC
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 09:59:44 +0000 (18:59 +0900)]
[enco] Cleanup caffe test framework (#3856)
This commit introduce TESTCASE_BASE_DIR variable which indicates where
testcases are located.
This is a preliminary work for shared caffe testcases.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 09:56:52 +0000 (18:56 +0900)]
[nnkit.support.tf] Remove dtype calls (#3854)
As a step to decouple tftestinfo from TensorFlow, this commit removes
dtype method calls from nnkit.support.tf.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 09:41:28 +0000 (18:41 +0900)]
[moco/tf] Addding padding data annotation into Conv2D (#3853)
This commit adds padding data annotation into Conv2D, which will be used by transforms after graph building.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 09:35:33 +0000 (18:35 +0900)]
[tfkit] Introduce convert command (#3852)
This commit extends tfkit with convert command. The current
implementation supports only conversion from pbtxt into json.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 07:34:34 +0000 (16:34 +0900)]
[tf2tflit] Add TF_IV3_AvgPool_GlobalPooling test (#3851)
This test corresponds to the last AvgPool node in inception v3 that
performs global average pooling.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 07:27:13 +0000 (16:27 +0900)]
[tf2tflite] Add TF_IV3_AvgPool_Module test (#3850)
This commit adds TF_IV3_AvgPool_Modeul test which covers all the AvgPool
nodes in IV3 except the last one.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 07:26:55 +0000 (16:26 +0900)]
[tf2tflite] Add TF_IV3_MaxPool test (#3847)
This commit adds TF_IV3_MaxPool test.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 07:21:39 +0000 (16:21 +0900)]
[moco/tf] fix shape for feature decode (#3842)
* [moco/tf] fix shape for feature decode
This will correct fix shape for feature decode, output as tensor from feature input
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix comment
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 07:10:56 +0000 (16:10 +0900)]
[moco/tf] fix shape for filter encode (#3844)
* [moco/tf] fix shape for filter encode
This will implement fix shape for filter encode, output as filter from tensor input
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix comment
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 06:55:21 +0000 (15:55 +0900)]
[moco/tf] fix shape for feature encode (#3843)
This will correct fix shape for feature encode, output as feature from tensor input
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 06:48:17 +0000 (15:48 +0900)]
[tf2tflite] Sync .pbtxt and .info on change (#3838)
The current implementation does not reflect the changes on .pbtxt and
.info.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 06:47:34 +0000 (15:47 +0900)]
Introduce nncc_find_resource (#3823)
* Introduce nncc_find_resource
This commit introduces nncc_find_resource which allows us to share the
common resource files across various modules.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Use NAME instead of PREFIX
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 06:01:18 +0000 (15:01 +0900)]
[i5diff] Add missing license (#3839)
This commit adds missing license notice.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 18 Jun 2019 06:00:26 +0000 (15:00 +0900)]
[loco exporter] Support FilterEncode (#3845)
This commit supports FilterEncode node for loco exporter
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 05:57:34 +0000 (14:57 +0900)]
[tf2tflite] Add MaxPool 000 testcase (#3827)
* [tf2tflite] Add MaxPool 000 testcase
This commit adds MaxPool 000 testcase which checks whether tf2tflite is
able to convert a simple tf model only with one MaxPool operation.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Fix shape
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 05:11:55 +0000 (14:11 +0900)]
[moco/tf] tensor and filter for ShapeInferenceData (#3837)
This will add getter of tensor and setter of filter for ShapeInferenceData
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 04:08:01 +0000 (13:08 +0900)]
[tf2tflite] Use i5diff to detect shape mismatch (#3836)
It turns out that h5diff exited with zero even in the presence of
shape mismatch.
This commit updates tf2tflite test framework to use i5diff (instead of
h5diff) to detect such shape mismatch cases.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 18 Jun 2019 03:50:17 +0000 (12:50 +0900)]
[loco exporter] Refactor export Transpose (#3835)
FeatureEncode, FeatureDecode are now exported as TRANSPOSE in tflite
when they are not identity(NHWC). In future, FilterEncode will be as
well. This commit refactors exporting TRANSPOSE in a form reusable
by FilterEncode.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 18 Jun 2019 03:45:15 +0000 (12:45 +0900)]
[locomotiv] Session setting input may throw error (#3714)
* [locomotiv] Session setting input may throw error
Previously Session::set_input() returns false on undesired cases. With
this commit, set_input() throws runtime error on such cases.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Review feedback: amend comment
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 03:20:47 +0000 (12:20 +0900)]
Introduce i5diff (#3832)
* Introduce i5diff
This commit introduces i5diff which takes two HDF5 files (that
nnkit HDF5 exports generates) and compares them.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Enable format check
* Fix typos
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 03:03:18 +0000 (12:03 +0900)]
[moco/tf] remove unnecessary call to toupper (#3834)
This will remove unnecessary call to toupper of pad type string as it is already uppercase when created
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 18 Jun 2019 03:00:28 +0000 (12:00 +0900)]
[loco] Enable Conv2D pad setter (#3833)
This commit enable (temporarily disabled) Conv2D pad setter.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 02:25:35 +0000 (11:25 +0900)]
[moco/tf] add log for fix padding avgpool2d (#3829)
* [moco/tf] add log for fix padding avgpool2d
This will add result pad values of fix padding of AvgPool2D node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* move LOGGER to top of function
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 02:23:31 +0000 (11:23 +0900)]
[moco/tf] Adding BiasAdd (#3815)
[moco/tf] Adding BiasAdd
This comits adds BiasAdd for moco/tf.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
Signed-off-by: Pavel Iliutchenko p.iliutchenk@samsung.com
Ivan Vagin/AI Tools Lab /SRR/Engineer/삼성전자 [Tue, 18 Jun 2019 01:23:09 +0000 (04:23 +0300)]
[moco/tf] provided helper for validate methods (#3804)
* [moco/tf] provided helper for frontend validate method
provided helper for frontend validate method
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
* Fixed formatting
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 01:17:49 +0000 (10:17 +0900)]
[moco/tf] Fix padding for MaxPool2D (#3805)
* [moco/tf] Fix padding for MaxPool2D
This will fix padding for MaxPool2D: same as AvgPool2D
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix formula
* fix pad formula
* fix typo
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 18 Jun 2019 01:09:01 +0000 (10:09 +0900)]
[moco/tf] test of shape inference for avgpool2d (#3825)
This wll add test of TestShapeTransform AvgPool2D for SAME and VALID padding type
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 17 Jun 2019 23:23:18 +0000 (08:23 +0900)]
[loco exporter] Support AvgPool2D (#3820)
This commit supports AvgPool2D loco node.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 17 Jun 2019 22:32:53 +0000 (07:32 +0900)]
[angkor] Introduce README.md (#3331)
* [angkor] Introduce README.md
This commit will add README file for `angkor`
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
* remove line
* Add words
* Fix headlines
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 17 Jun 2019 11:20:01 +0000 (20:20 +0900)]
[tf2tflite] Ignore name mismatch on testing (#3826)
tfttflite now does not count tensor name mismatch as a test failure.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 17 Jun 2019 11:16:35 +0000 (20:16 +0900)]
[moco/tf] Test pbtxt extracted from inception_v3 (#3824)
These test pbtxt files are extracted from inception_v3, containing 4, 7, 10 nodes respectively.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 17 Jun 2019 10:44:28 +0000 (19:44 +0900)]
[moco/tf] remove unnecessary comment (#3822)
This will remove unnecessary comment about padding
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 17 Jun 2019 10:42:30 +0000 (19:42 +0900)]
[nnkit-HDF5] Use index instead of name (#3821)
Previously, nnkit HDF5 export uses a tensor name when it create a
dataset. With this commit, nnkit HDF5 export uses a tensor index
instead.
This commit also updates nnkit HDF5 import action accordingly.
To maintain backward compatibility, HDF5 export stores the name of
each tensor as an independent attribute. This additional attribute
allows users to detect name mismatch same as before.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 17 Jun 2019 10:26:10 +0000 (19:26 +0900)]
[moco/tf] store padding case insensitive (#3819)
This will change to store padding string as case insensitive as it will be stored once and used many
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 17 Jun 2019 10:23:23 +0000 (19:23 +0900)]
[moco/tf] fix padding for avgpool2d (#3816)
* [moco/tf] fix padding for avgpool2d
This will fox padding calculation formula for AvgPool2D
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* with dilation and log