윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 8 Jul 2019 07:10:50 +0000 (16:10 +0900)]
[locoex/customop] Initial commit for Custom Op (#4136)
* [locoex/customop] Initial commit for Custom Op
This is an initial commit for Custom Op. Parents for Custom Op will be added after successive PRs.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* revised
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 8 Jul 2019 02:45:27 +0000 (11:45 +0900)]
[moco/tf] Introduce ResolveFusedBatchNorm (#4125)
This will introduce ResolveFusedBatchNorm transformation that decomposes FusedBatchNorm node into Add and Mul
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 8 Jul 2019 01:50:37 +0000 (10:50 +0900)]
[exo.tflite] Extract Type Inference as a separate phase (#4126)
This commit extracts (Data) Type Inference from exportOpDefinedTensor
phase.
This chanage allows users to infer types without exporting tensors.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 8 Jul 2019 01:11:39 +0000 (10:11 +0900)]
[plier/tf] Move more functions from Moco::tf::convert.h & cpp (#4122)
The following was moved into plier.tf from moco/../convert.h & cpp
- two functions
- test case
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 8 Jul 2019 01:06:31 +0000 (10:06 +0900)]
[moco/tf] name attribute for transforms (#4121)
This will add name() attribute for existing Transforms
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 8 Jul 2019 00:55:21 +0000 (09:55 +0900)]
[moco/tf] simplify get_data_layout (#4124)
This will simplify get_data_layout to call as_DataLayout as common part
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 8 Jul 2019 00:14:54 +0000 (09:14 +0900)]
Add "GoogleDoubleConversion" package (#4112)
* Add "GoogleDoubleConversion" package
This commit introduces "GoogleDoubleConversion" which allows users to
import system-wide Google DoubleConversion library in nncc style.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Remove tab
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 8 Jul 2019 00:14:32 +0000 (09:14 +0900)]
[exo/tflite] Use loco::output_nodes in exportOpDefinedTensors (#4118)
exportOpDefinedTensors currently uses its own implementation of
loco::output_nodes.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 8 Jul 2019 00:11:36 +0000 (09:11 +0900)]
[moco_tf] Fix BiasAdd test indent todo (#4120)
This will fix indentation of BiasAdd test from previous change
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 8 Jul 2019 00:08:21 +0000 (09:08 +0900)]
[moco/tf] Conv2D test support TFConv2D (#4083)
* [moco/tf] Conv2D test support TFConv2D
This will split Conv2D test to support TFConv2D by the knob
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* include knob
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 7 Jul 2019 23:29:51 +0000 (08:29 +0900)]
[moco/tf] Rename FusedBatchNorm Graph builder (#4119)
This will rename GraphBuilder for FusedBatchNorm with 'TF' prefix to match other GraphBuilders
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 5 Jul 2019 08:49:35 +0000 (17:49 +0900)]
[moco/tf] Add Canonicalization and Knob for TFBiasAdd (#4117)
* [moco/tf] Add Canonicalization and Knob for TFBiasAdd
This will Add Canonicalization class with TFBiassAdd node.
Add a Knob for canonicalization of TFBiasAdd to Canonical BiasAdd node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix note
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 5 Jul 2019 08:47:43 +0000 (17:47 +0900)]
[loco] Introduce FeatureBiasAdd node (#4110)
This commit introduces FeatureBiasAdd which adds a "bias" along
a FeatureMap's depth axis.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 5 Jul 2019 08:25:26 +0000 (17:25 +0900)]
[moco/tf] Import as TFBiasAdd by Knob (#4116)
This will add ImportAsTFBiasAdd knob and import as TFBiasAdd when turned on.
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 5 Jul 2019 07:51:28 +0000 (16:51 +0900)]
[moco/tf] Move helper alias to TFNodeDecl header (#4115)
This will move helper alias from TFConv to TFNodeDecl header for common usage
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 5 Jul 2019 07:33:24 +0000 (16:33 +0900)]
[moco/tf] Introduce TFBiasAdd (#4114)
This will introduce TFBiasAdd to import BiasAdd node as-is
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 5 Jul 2019 05:24:38 +0000 (14:24 +0900)]
[exo/tflite] Extract Type/Shape Context (#4107)
SerializedModelData has multiple responsibilities as of now.
This commit extracts TypeContext (which manages node-to-type mapping)
and ShapeContext (which manages node-to-shape mapping) as the first step
to internal design revision.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 5 Jul 2019 04:08:10 +0000 (13:08 +0900)]
Add TF_SMALL_NET_0006 test (#4103)
* Add TF_SMALL_NET_0006 test
This test checkes what happens when "Concat" is in between two Feature operations.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Update comment
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 5 Jul 2019 04:07:59 +0000 (13:07 +0900)]
[moco/tf] Set Graph Input/Output Data Type (#4104)
From now on, moco::tf::Importer generates a loco graph with
valid Input/Output dtype values.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 5 Jul 2019 04:07:49 +0000 (13:07 +0900)]
[tf2tflite] Refine how to add tests (#4105)
This commit refines tf2tflite test framework to use CMake's "include"
command instead of ListFile_Read function.
This change permits flexible test management. For example, now it is
possible to comment out some tests temporarily.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 5 Jul 2019 01:15:41 +0000 (10:15 +0900)]
[loco] Introduce TensorBiasAdd alias (#4102)
This change allows users to use a bit short "loco::TensorBiasAdd" instead
of "loco::BiasAdd<loco::Domain::Tensor>".
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 5 Jul 2019 01:04:45 +0000 (10:04 +0900)]
[mocotest/onnx] Generate test file only on change (#4106)
The current implementation always copies and generates test files. With
this commit, test files are copied and generated only when it is
necessary.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 4 Jul 2019 22:05:27 +0000 (07:05 +0900)]
[TFtest] Add TF_SMALL_NET_0004 test (#4095)
* [TFtest] Add TF_SMALL_NET_0004 test
This commit adds TF_SMALL_NET_0004 test which sequentially performs
DepthwiseConv2d - FusedBatchNorm
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
* unpack values
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 10:27:03 +0000 (19:27 +0900)]
[loco] Add DataType attribute to Graph Input/Output (#4097)
This commit allows loco frontends to record the data type of graph-level
input/output in GraphInput/GraphOutput.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 4 Jul 2019 10:11:46 +0000 (19:11 +0900)]
[locomotiv] Introduce ReLU6 operation to locomotiv (#4096)
This commit introduce ReLU6 operation to locomotiv.
Signed-off-by: seongwoo <sw4670.chae@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 09:22:37 +0000 (18:22 +0900)]
[mocotest/tf] Copy files only on change (#4094)
moco testframework now copies (and generates) test files only on change.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 4 Jul 2019 07:45:55 +0000 (16:45 +0900)]
[TFtest] Add TF_SMALL_NET_0003 test (#4093)
This commit adds TF_SMALL_NET_0003 test which sequentially performs
Conv2d - FusedBatchNorm
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 07:31:04 +0000 (16:31 +0900)]
[mocotest/tf] Use shared testcases (#4091)
mocotest-tf now uses shared TensorFlow tests.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 07:29:25 +0000 (16:29 +0900)]
[tf2tflite] Use shared testcases (#4090)
This commit rewrites tf2tflite test framework to use shared testcases
and removes tf2tflite-local testcases.
tf2tflite test framework now manages whitelist internally.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 4 Jul 2019 04:49:14 +0000 (13:49 +0900)]
[mocotest-tf] test case for custom op (#4078)
* [mocotest-tf] test case for custom op
This is the first testcase for custom op. code for moco/loco extension will be written to pass this test.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* move into new resource dir
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 04:44:29 +0000 (13:44 +0900)]
[moco/tf] Implement ReorderDecode Transform (#4082)
* [moco/tf] Implement ReorderDecode Transform
This commit implements ReorderDecode transform which rewrites Decode ->
Op as Op -> Decode if possible.
The current implementation works only for ReLU.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Hoist loop-invariant code
* Update Graph example
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 04:43:33 +0000 (13:43 +0900)]
[angkor] Add TensorIndex & TensorShape alias (#4085)
This commit introduces an alias of tensor::Index and tensor::Shape under
nncc::core::ADT namespace (to shorten client-side code).
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 04:42:15 +0000 (13:42 +0900)]
[tf2tflite] Detect crashes during test (#4089)
* [tf2tflite] Detect crashes during test
The current test frameworks becomes green as long as i5diff succeeds,
and thus it cannot detect any crash in tf2tflite / nnkit-run.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Remove trap code
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 4 Jul 2019 03:59:10 +0000 (12:59 +0900)]
[moco/tf] Introduce TFConv2D IR (#4068)
* [moco/tf] Introduce TFConv2D IR
This will introduce TFConv2D TensorFlow dialect for Conv2D and related changes
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use reference
* as type
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 01:47:42 +0000 (10:47 +0900)]
Introduce ListFile CMake module (#4081)
This commit introduces ListFile CMake module which provides
ListFile_Read helper.
This helper makes it easy to create a list from text files.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 4 Jul 2019 01:36:56 +0000 (10:36 +0900)]
Introduce TensorFlowTests resource (#4080)
This commit introduces TensorFlowTests resource which includes all the
tests in moco-tf and tf2tflite.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 4 Jul 2019 01:29:52 +0000 (10:29 +0900)]
[moco/tf] Branch Conv2D graph builder for TFDialect (#4079)
* [moco/tf] Branch Conv2D graph builder for TFDialect
This will prepare Conv2D graph builder with a knob to generate TFConvD node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use throw
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 4 Jul 2019 00:15:40 +0000 (09:15 +0900)]
[moco_tf] tensor_shape setter in ShapeInferenceData (#4084)
This will add tensor_shape setter in ShapeInferenceData
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 3 Jul 2019 23:52:39 +0000 (08:52 +0900)]
[moco/tf] Add more methods in Convert (#4070)
* [moco/tf] Add more methods in Convert
This will add more methods in Convert for import
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use ASSERT_EQ
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 3 Jul 2019 09:12:48 +0000 (18:12 +0900)]
[loco] ReLU6 node (#4077)
This commit introduce ReLU6 node to loco.
Signed-off-by: seongwoo <sw4670.chae@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 3 Jul 2019 08:16:20 +0000 (17:16 +0900)]
[mocotest/tf] Use TEST_REPO & TEST_SPACE variables (#4074)
This commit updates mocotest-tf's CMakeLists.txt to use TEST_REPO &
TEST_SPACE variables instead of CMAKE_CURRENT_SOURCE_DIR &
CMAKE_CURRENT_BINARY_DIR.
This change gives some flexibility on test directory layout.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 3 Jul 2019 07:56:33 +0000 (16:56 +0900)]
[moco/tf] Run TFOptimizier and Canonicalizer (#4073)
This will make frontend to run TFOptimizier and Canonicalizer
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 3 Jul 2019 07:56:03 +0000 (16:56 +0900)]
[moco/tf] Knob for ResolveFusedBatchNorm (#4075)
This will add a knob for ResolveFusedBatchNorm transform
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 3 Jul 2019 07:29:42 +0000 (16:29 +0900)]
[plier/tf] copy_shape() function (#4066)
* [plier/tf] copy_shape() function
This adds copy_shape() function that copies tensorflow shape proto to angkor shape.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* comment
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 3 Jul 2019 07:28:43 +0000 (16:28 +0900)]
[tf2tflite] Use exo-tflite instead of loco-exporter (#4071)
This commit revises tf2tflite to use exo-tflite instead of
loco-exporter. This change resolves conflicts when loco-exporter
evolves into a in-house package exporter.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 3 Jul 2019 06:31:45 +0000 (15:31 +0900)]
[locomotiv] Add Reshape operation (#4033)
* [locomotiv] Add Reshape operation
This commit introduce Reshape operation to locomotiv.
Signed-off-by: seongwoo <sw4670.chae@samsung.com>
* This commit modify code according to comments.
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 3 Jul 2019 06:31:31 +0000 (15:31 +0900)]
Introduce exo-tflite (#4050)
* Introduce exo-tflite
This commit introduces exo-tflite project. All the code itself originates
from loco-exporter, but have small style changes.
This "exo-tflite", unlike loco-exporter, will not evolve into an exporter for
in-house package.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Fix ifdef guard
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 3 Jul 2019 05:42:43 +0000 (14:42 +0900)]
[moco/tf] Introduce Canonicalizer (#4045)
* [moco/tf] Introduce Canonicalizer
This will introduce Canonicalizer for transforming TFNodes to CanonicalNodes
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* remove test
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 3 Jul 2019 05:00:55 +0000 (14:00 +0900)]
[tf2tflite] Rename testcases (#4064)
This commit renames the tests that have a name conflict with moco-tf.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 3 Jul 2019 05:00:41 +0000 (14:00 +0900)]
[loco] Support FeatureCodec clone (#4062)
This commit introduce "clone" API to FeatureEncoder/Decoder interface.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 3 Jul 2019 04:20:02 +0000 (13:20 +0900)]
Introduce NNCC_EXTERNALS_DIR configuration (#4065)
This commit allows users to download external dependencies
into their own location via setting NNCC_EXTERNALS_DIR.
Note that the default behavior is same as before.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 3 Jul 2019 03:16:35 +0000 (12:16 +0900)]
[plier/tf] Introducing plier/tf (#4056)
* [plier/tf] Introducing plier/tf
This commit introduce `plier/tf`, which is a set of small tools to handle Tensorflow model. These are currently copied from `moco/../Convert.h and cpp` and `moco/../TestHelper.h and cpp`.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* typo
* remove loco
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 3 Jul 2019 01:40:15 +0000 (10:40 +0900)]
[moco/tf] Introduce TFMul IR (#4061)
This will add TFMul IR and related codes to support TensorFlow Mul node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 3 Jul 2019 00:22:56 +0000 (09:22 +0900)]
[moco/tf] Add transforms for TFAdd (#4060)
This will add missing TFAdd related transforms
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 23:59:59 +0000 (08:59 +0900)]
[moco/tf] Import FusedBatchNorm (#4053)
* [moco/tf] Import FusedBatchNorm
This will add import of FusedBatchNorm to TFFusedBatchNorm IR
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* add required transformation handlers
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 23:59:02 +0000 (08:59 +0900)]
[moco/tf] Test for TFFusdBatchNorm (#4052)
This will add ctor unit test of TFFusedBatchNorm
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 22:26:17 +0000 (07:26 +0900)]
[moco/tf] Introduce TFAdd IR (#4051)
* [moco/tf] Introduce TFAdd IR
This will introduce TensorFlow dialect Add IR
- this will be used as intermediate node while resolving FusedBatchNorm node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* apply comments
* another comment
* privode GraphDef example
이성재/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 09:21:47 +0000 (18:21 +0900)]
Update README.modified (#4048)
- Add `Mainterners` and `Committers` list to following the rule of Samsung
Inner Source Program.
Signed-off-by: Sung-Jae Lee <sj925.lee@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 2 Jul 2019 08:03:04 +0000 (17:03 +0900)]
[tf2tflite] Add TF_SMALL_NET_0001 test (#4047)
This commit adds TF_SMALL_NET_0001 test which sequentially performs
Conv2D-BiasAdd-Conv2D operations.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 07:45:38 +0000 (16:45 +0900)]
[moco/tf] Introduce TFOptimizier (#4049)
This will introduce TFOptimizier to provide transformations with TensorFlow Dialects
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 07:43:35 +0000 (16:43 +0900)]
[moco/tf] Add ImportAsTFConv2D knob (#4044)
This will add ImportAsTFConv2D in knob to switch Conv2D import as Canonical or TensorFlow Conv2D
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 2 Jul 2019 05:03:06 +0000 (14:03 +0900)]
Add TensorFlow 1.14 source package (#4043)
This commit introduces TensorFlow v1.1.4 source package which allows
each sub-project to access T/F 1.14 source during build.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 04:30:17 +0000 (13:30 +0900)]
[mocotest/tf] FusedBatchNorm_001 (#4042)
This will add FusedBatchNorm_001 test material having values for calculation validation
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 04:29:48 +0000 (13:29 +0900)]
[mocotest/tf] FusedBatchNorm_000 (#4041)
This will add FusedBatchNorm_000 test material for minimul testing for FusedBatchNorm node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 2 Jul 2019 04:05:52 +0000 (13:05 +0900)]
[moco/tf] Introduce Domain Conversion Simplification (#4037)
This commit introduces Domain Conversion Simplification optimizations
which convertes FeatureDecode followed by FeatureEncode as Forward.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 2 Jul 2019 03:12:59 +0000 (12:12 +0900)]
[moco/tf] Move Transform.h to the top-level (#4038)
This change makes it easy to distinguish Transform declaration (Transform.h) and
enumerate its implementations (all the files under Transforms/).
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 03:05:56 +0000 (12:05 +0900)]
[moco/tf] Introduce TFFusedBatchNorm IR (#4039)
This will introduce TFFusedBatchNorm to import TensorFlow FusedBatchNorm node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 2 Jul 2019 00:51:15 +0000 (09:51 +0900)]
[moco/tf] Add TFNode visitor (#4036)
* [moco/tf] Add TFNode visitor
This will add visitor related methods in TFNode
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix comments
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 1 Jul 2019 10:50:32 +0000 (19:50 +0900)]
[moco/tf] Implement Dead Node Elimination (#4032)
Now, moco.tf compilation pipiline support Dead Node Elimination. Note that
this optimization is off by default.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 09:31:33 +0000 (18:31 +0900)]
[moco/tf] Introduce TFNode for TensorFlow dialect (#4029)
This will introduce TFNode and TFNodeImpl for Tensorflow dialect
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 1 Jul 2019 09:03:40 +0000 (18:03 +0900)]
[loco] Introduce active_nodes helper (#4014)
This commit introduces active_nodes helpers which returns all the
predecessor of a given set of nodes.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 07:45:06 +0000 (16:45 +0900)]
[moco/tf] Rename macro as CANONICAL_NODE (#4025)
* [moco/tf] Rename macro as CANONICAL_NODE
This will rename existing MOCONODE to CANONICAL_NODE as they are actually treating loco canonical nodes
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use canonical
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 07:12:05 +0000 (16:12 +0900)]
[moco/tf] Introduce TFOpcode (#4028)
This will introduce TFOpcode for TensorFlow dialect
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 1 Jul 2019 06:41:02 +0000 (15:41 +0900)]
[moco/tf] Implement forward node elimination (#4027)
This commit implements forward node elimination as a part of moco.tf
compilation pipeline.
Note that this transform is off by default.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 05:23:37 +0000 (14:23 +0900)]
[moco/tf] Import to Importer (#4019)
* [moco/tf] Import to Importer
This will rename Import class to Importer
- abd method load() to `import`
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix header
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 05:13:10 +0000 (14:13 +0900)]
[moco_tf] Introduce TFDialect (#4020)
* [moco_tf] Introduce TFDialect
This will introduce TFDialect inherited from loco::Dialect to provide IRs for TensorFlow dialect.
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* rename to TFDialect
* fix typo
* fix rename
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 05:09:48 +0000 (14:09 +0900)]
[moco/tf] Use TensorFlow (#4026)
This will fix comments using Tensorflow to TensorFlow
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 04:25:43 +0000 (13:25 +0900)]
[mocotest] Test for Mul (#4024)
This will introduce test for Mul node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 04:25:11 +0000 (13:25 +0900)]
[mocotest] Test for Add (#4023)
This will introduce Test for Add Node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 1 Jul 2019 04:08:55 +0000 (13:08 +0900)]
[tf2tflite] Pass test configurations via .config file (#4017)
* [tf2tflite] Pass test configurationss via .config file
Now, "testall" script takes test configurations via .config
file, which makes it easy to re-invoke tests from command-line.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Update comment to reflect the actual implementation
이성재/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 1 Jul 2019 03:52:44 +0000 (12:52 +0900)]
Add CONTRIBTUTORS list (#4021)
- Add contributor list, as is.
- Generated by `git shortlog -se | cut -f2` with `.mailmap` file.
Signed-off-by: Sung-Jae Lee <sj925.lee@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 1 Jul 2019 03:42:24 +0000 (12:42 +0900)]
[moco.tf] Introduce Optimizer (#4013)
This commit introduces Optimizer class and updates Frontend to invoke
this optimization after import.
This Optimizer will performs dedicated transforms on imported graphs,
but the current implementation serves just as a placeholder and has
no transforms yet.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Sun, 30 Jun 2019 23:27:22 +0000 (08:27 +0900)]
[loco] modify struct to class (#4016)
This commit modify struct to class in Nodes.h for unity
Signed-off-by: seongwoo <sw4670.chae@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 28 Jun 2019 04:29:58 +0000 (13:29 +0900)]
[tf2tflite] Add TF_SMALL_NET_0000 test (#4012)
This commit adds TF_SMALL_NET_0000 test which includes a simple network
consists of Conv2D and MaxPool.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 28 Jun 2019 04:29:20 +0000 (13:29 +0900)]
[loco] Check references on node destruction (#4010)
This commit adds an assert to check references in Node destructor in
order to detect dangling references as early as possible.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 28 Jun 2019 04:02:53 +0000 (13:02 +0900)]
[tf2tflite] Support logging (#4009)
Now, users are able to enable tf2tflite logging by setting TF2TFLITE_Log
environment variable as a non-zero value.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 28 Jun 2019 01:46:55 +0000 (10:46 +0900)]
[loco] Support node destruction (#4006)
Some optimizations (e.g. dead node elimination) needs node destruction, but
it is currently impossible to destroy nodes.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 28 Jun 2019 01:29:08 +0000 (10:29 +0900)]
[moco.tf] Extract Phase infrastructure (#4008)
This commit extracts "transform" loop in Import as reusable Phase
class.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 28 Jun 2019 01:25:38 +0000 (10:25 +0900)]
[loco] Introduce CanonicalNodeDecl.h (#4007)
The current implementation requires users to include CanonicalNodeImpl.h
additionally.
To eliminate such burden, this commit introduces CanonicalNodeDecl.h and
revises CanonicalNode.h to includes both decl/impl headers.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 27 Jun 2019 07:14:49 +0000 (16:14 +0900)]
[loco] Introduce Node.opnum API (#4004)
This opnum API allows us to check the type of nodes without using
dynamic_cast.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 27 Jun 2019 07:10:32 +0000 (16:10 +0900)]
[locop] Support TensorConcat (#4001)
TensorConcat now has its own pretty printer.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 27 Jun 2019 07:10:18 +0000 (16:10 +0900)]
[moco/tf] How to convert TensorFlow Const (#4002)
Conversion.md now explains how to convert T/F Const node (which
tf.constant API internally generates) into loco.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 27 Jun 2019 06:04:41 +0000 (15:04 +0900)]
[loco] Require Node to implement dialect method (#3999)
* [loco] Require Node to implement dialect method
The current design provides a default implementation. This commit
removes this default implementation of dialect method and requires each
node to implement it.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Return const Dialect *
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 27 Jun 2019 04:32:27 +0000 (13:32 +0900)]
[hermes/std] Introduce EnvConfig (#4000)
This commit introduces a reusable EnvConfig implementation.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 27 Jun 2019 03:05:14 +0000 (12:05 +0900)]
[moco_tf] Introduce common import in Frontend (#3992)
* [moco_tf] Introduce common import in Frontend
This will introduce private import method in Frontend as common of load
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* include header
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 27 Jun 2019 02:22:19 +0000 (11:22 +0900)]
Extract nnkit-misc from nnkit (#3996)
Let's extract nnkit-misc that implements various helpers for writing
nnkit extensions.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 27 Jun 2019 01:24:02 +0000 (10:24 +0900)]
[moco-tf] Revisit tests as Import (#3991)
This will revisit unit tests as Import and do Import testing
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 26 Jun 2019 23:44:34 +0000 (08:44 +0900)]
[moco/tf] convert pb bool attribute to C++ bool (#3995)
This commit adds get_bool_attr(..) to access tensorflow bool attr
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 26 Jun 2019 22:11:31 +0000 (07:11 +0900)]
[moco.log] Reflect hermes design change (#3994)
This commit updates moco.log implementations according to the recent
hermes design update.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 26 Jun 2019 10:41:11 +0000 (19:41 +0900)]
Extract nnkit-onnxrt from nnkit (#3989)
Let's allow each module to depend on nnkit-onnxrt.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>