platform/core/ml/nnfw.git
4 years ago[exo] Turning on knob to enabling fusion (of add/sub, relu/relu6) by default (#8333)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 06:30:36 +0000 (15:30 +0900)]
[exo] Turning on knob to enabling fusion (of add/sub, relu/relu6) by default (#8333)

This turns on knobs to enabling fusion (of add/sub, relu/relu6) by default

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[loco] DepthwiseFilterDecode canonical IR (#8332)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 21 Oct 2019 06:30:26 +0000 (15:30 +0900)]
[loco] DepthwiseFilterDecode canonical IR (#8332)

This commit introduces DepthwiseFilterDecode loco canonical IR

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[tf2tflite] added NET_0035-37 (CBR related test) into test.lst (#8343)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 06:26:13 +0000 (15:26 +0900)]
[tf2tflite] added NET_0035-37 (CBR related test) into test.lst (#8343)

This adds NET00_35, NET_0036, NET_0037 (CBR related test) into test.lst

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[exo] DepthwiseConv2D TFL dialect IR (#8320)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 21 Oct 2019 06:24:42 +0000 (15:24 +0900)]
[exo] DepthwiseConv2D TFL dialect IR (#8320)

* [exo] DepthwiseConv2D TFL dialect IR

This commit introduces TFLDepthwiseConv2D, a TFL dialect IR for
DepthwiseConv2D. It also introduces its summary.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Add default value and test

4 years ago[moco-integration-test] adding NET_0035, NET_0037 into test.lst (#8346)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 06:21:57 +0000 (15:21 +0900)]
[moco-integration-test] adding NET_0035, NET_0037 into test.lst (#8346)

This adds NET_0035, NET_0037 into test.lst.
0036 fails. An issue will be created.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[nnfw-api] Merge custom op API implementation (#8340)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 21 Oct 2019 05:59:21 +0000 (14:59 +0900)]
[nnfw-api] Merge custom op API implementation (#8340)

Move custom op API implementation into neurun/api

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years agoUpdate BuildGraph parameter (#8337)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 21 Oct 2019 05:45:08 +0000 (14:45 +0900)]
Update BuildGraph parameter (#8337)

Current BuildGraph get subgraph parameter, not interpreter

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years ago[moco] Import Nodes (#8336)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 05:26:59 +0000 (14:26 +0900)]
[moco] Import Nodes (#8336)

This will add Import GraphBuilder Nodes to import TensorFlow NodeDef

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[loco] PemutingDecoder for DepthwiseFilter (#8331)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 21 Oct 2019 05:26:24 +0000 (14:26 +0900)]
[loco] PemutingDecoder for DepthwiseFilter (#8331)

* [loco] PemutingDecoder for DepthwiseFilter

This commit introduces PemutingDecoder for DepthwiseFilter

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Remove unrelated

4 years ago[moco-tf] use mio-tf (#8328)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 04:48:58 +0000 (13:48 +0900)]
[moco-tf] use mio-tf (#8328)

* [moco-tf] use mio-tf

This will let moco-tf to use mio-tf so that one TensorFlow protobuf library produced and used

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* check mio_tf cond

4 years ago[moco] Import requires (#8339)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 04:46:59 +0000 (13:46 +0900)]
[moco] Import requires (#8339)

This will update Import requires file

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[logo] Update SimplifyDomainConversionPass comment (#8323)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 21 Oct 2019 03:13:46 +0000 (12:13 +0900)]
[logo] Update SimplifyDomainConversionPass comment (#8323)

This commit updates SimplifyDomainConversionPass comment to cover its
whole functionality

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[exo] Update make_filter_* comment (#8322)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 21 Oct 2019 03:12:43 +0000 (12:12 +0900)]
[exo] Update make_filter_* comment (#8322)

This commit fixes typo and updates following helpers in GraphBlock.h:
- make_filter_encode
- make_filter_decode

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[plier-tf] use mio-tf (#8329)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 03:11:05 +0000 (12:11 +0900)]
[plier-tf] use mio-tf (#8329)

* [plier-tf] use mio-tf

This will let plier-tf to use mio-tf so that one TensorFlow protobuf library is produced

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* check mio_tf cond

* match cmake if endif

4 years ago[moco] Convert and TestHelper (#8330)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 03:09:08 +0000 (12:09 +0900)]
[moco] Convert and TestHelper (#8330)

This will introduce Convert for Import and TestHelper for Import testing

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[nnpkg, circle] Define instance_norm in circle_schema.fbs (#8262)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 02:36:52 +0000 (11:36 +0900)]
[nnpkg, circle] Define instance_norm in circle_schema.fbs (#8262)

It introduces instance_norm operator in circle schema.
Also, it updates circle_schema_generated.h for circle_loader.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
4 years ago[moco-tf] Redirect GraphBuilder to moco (#8327)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 01:59:19 +0000 (10:59 +0900)]
[moco-tf] Redirect GraphBuilder to moco (#8327)

This will update GraphBuilder and GraphBuilderContext to redirect to moco for migration

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years agoRelocate compute libraries (#8307)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 21 Oct 2019 01:45:22 +0000 (10:45 +0900)]
Relocate compute libraries (#8307)

Rename srcn to ncnn
Relocate compute libraries into compute/ (armcompute-ex, cker, ncnn)

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years ago[loco] DepthwiseFilterDecoder (#8319)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 21 Oct 2019 01:43:12 +0000 (10:43 +0900)]
[loco] DepthwiseFilterDecoder (#8319)

This commit introduces DepthwiseFilterDecoder class.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[moco] Introduce TFPlaceholder IR (#8313)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 21 Oct 2019 01:27:27 +0000 (10:27 +0900)]
[moco] Introduce TFPlaceholder IR (#8313)

* [moco] Introduce TFPlaceholder IR

This will introduce TFPlaceholder IR for Placeholder node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* add todo comment

4 years agoSupport Logistic in base loader (#7942)
이춘석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 21 Oct 2019 01:01:10 +0000 (10:01 +0900)]
Support Logistic in base loader (#7942)

- Add logistic operation on base loader

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
4 years ago[loco] Fix typo for DepthwiseFilterEncoder comment (#8318)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Sun, 20 Oct 2019 21:24:46 +0000 (06:24 +0900)]
[loco] Fix typo for DepthwiseFilterEncoder comment (#8318)

This commit fixes typo in DepthwiseFilterEncode comment

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[moco] Include moco import (#8316)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 20 Oct 2019 21:19:23 +0000 (06:19 +0900)]
[moco] Include moco import (#8316)

This will update CMake to include import as sub project

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[mir] Refine operations constructors. (#8283)
Gusev Dmitry/Engineer/AI Tools Lab /SRR/Samsung Electronics [Fri, 18 Oct 2019 20:34:43 +0000 (23:34 +0300)]
[mir] Refine operations constructors. (#8283)

Removed obsolete operations constructors in [mir].

Signed-off-by: Dmitry Gusev <d.gusev@partner.samsung.com>
4 years ago[mir] Remove Operation::Input class (#8294)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Fri, 18 Oct 2019 20:33:38 +0000 (23:33 +0300)]
[mir] Remove Operation::Input class (#8294)

Remove `Operation::Input` class and add lightweight `Use` class.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[mir] Remove unused functions (#8324)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Fri, 18 Oct 2019 19:21:24 +0000 (22:21 +0300)]
[mir] Remove unused functions (#8324)

Remove `replaceWithInputNode` and `replaceInputNodes` functions which were only used in unittests.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[mir] Introduce new fused bias constructors (#8186)
Pavel Iliutchenko/AI Tools Lab /SRR/Engineer/Samsung Electronics [Fri, 18 Oct 2019 10:12:29 +0000 (13:12 +0300)]
[mir] Introduce new fused bias constructors (#8186)

* Conv2DOp ctor (and fix copy)
* DepthwiseConv2DOp ctor (and fix copy)

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
4 years ago[base-loader] Change SubNode to MulNode in loadMul (#8311)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 09:37:30 +0000 (18:37 +0900)]
[base-loader] Change SubNode to MulNode in loadMul (#8311)

It fixes wrong code in loadMul.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
4 years ago[moco-tf] fix PadCanonicalizer (#8298)
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 18 Oct 2019 07:24:39 +0000 (16:24 +0900)]
[moco-tf] fix PadCanonicalizer (#8298)

* [moco-tf] fix PadCanonicalizer

This commit fixs PadCanonicalizer in moco-tf.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>
* add comment.

4 years ago[moco] Import: Introduce GraphBuilder (#8306)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 07:11:38 +0000 (16:11 +0900)]
[moco] Import: Introduce GraphBuilder (#8306)

This will introduce moco-import target with GraphBuilder and related classes copied from moco-tf

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[moco-tf] Refactor TF node graph builder test part 2 (#8297)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 07:11:17 +0000 (16:11 +0900)]
[moco-tf] Refactor TF node graph builder test part 2 (#8297)

This will refactor test NOT to use Importer module of node graph builder from MaxPool to Tanh

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[moco-tf] Refactor TF node graph builder test part 1 (#8296)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 07:10:48 +0000 (16:10 +0900)]
[moco-tf] Refactor TF node graph builder test part 1 (#8296)

This will refactor test NOT to use Importer module of node graph builder from Add to FusedBatchNorm

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[exo] replace PhaseStrategy of TestHelper.h from Saturate to Restart (#8304)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 05:08:20 +0000 (14:08 +0900)]
[exo] replace PhaseStrategy of TestHelper.h from Saturate to Restart (#8304)

This replaces PhaseStrategy of TestHelper.h from Saturate to Restart

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[res/TensorFlowTests] NET_0037 test case (#8302)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 05:07:29 +0000 (14:07 +0900)]
[res/TensorFlowTests] NET_0037 test case (#8302)

* [res/TensorFlowTests] NET_0037 test case

This adds NET_0037, Conv2D + ReLU + Add. By exo, Conv2D and ReLU are fused but Add isn't.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* simple description

4 years ago[nnfw-gbs] Merge build require and option setting (#8232)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 18 Oct 2019 04:28:40 +0000 (13:28 +0900)]
[nnfw-gbs]  Merge build require and option setting (#8232)

- Remove python3 build require
- Require libarmcl-devel and python on aarch target only
- Merge build option setting for coverage build, test build and general build

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years ago[exo] adding FuseReluPass into TFLOptimize.cpp (#8280)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 04:26:00 +0000 (13:26 +0900)]
[exo] adding FuseReluPass into TFLOptimize.cpp (#8280)

This adds FuseReluPass into TFLOptimize.cpp. When a knob is on, this pass will run.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[exo] rewriting a function to handle Mixin (#8281)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 04:23:12 +0000 (13:23 +0900)]
[exo] rewriting a function to handle Mixin (#8281)

Instread of trying to cast to a lot of type, code was re-written to try one-time cast to Mixin.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[lib/tflite] Update operation register (#8278)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 18 Oct 2019 04:19:49 +0000 (13:19 +0900)]
[lib/tflite] Update operation register (#8278)

Update register.h/cpp to match tflite v1.13.1
Support new operations introduce on v1.13.1
Support more upstream custom operation

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years ago[exo] FuseConv2DAddSubPass checks Conv2D's fused activation function (#8268)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 03:27:40 +0000 (12:27 +0900)]
[exo] FuseConv2DAddSubPass checks Conv2D's fused activation function (#8268)

* [exo] FuseConv2DAddSubPass checks Conv2D's fused activation function

This enables FuseConv2DAddSubPass to check Conv2D's fused activation function
and perform fusion considering the status of fused activation function.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* remove unused var

* test case

* one more test case

* float act_func(...) -> void act_func(..)

4 years ago[mio-tf] TensorFlow model library (#8274)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 18 Oct 2019 02:20:46 +0000 (11:20 +0900)]
[mio-tf] TensorFlow model library (#8274)

* [mio-tf] TensorFlow model library

This will introduce mio-tf as a separate project for accessing TensorFlow protobuf model file

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* merge cmake and apply comments

4 years ago[neurun] Backend priority by OP_BACKEND_ALLOPS (#8124)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 18 Oct 2019 02:09:37 +0000 (11:09 +0900)]
[neurun] Backend priority by OP_BACKEND_ALLOPS (#8124)

- Try backend loading from default backend setting, but not fail if backend is not exist
  - Fail if there is no loaded backend
  - Reorder to load acl(cl, neon) first
- If OP_BACKEND_ALLOPS is empty, use first loaded backend
- Set default OP_BACKEND_ALLOPS emptyy

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years agoRename moco-onnx as oneco (#8251)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 18 Oct 2019 00:48:40 +0000 (09:48 +0900)]
Rename moco-onnx as oneco (#8251)

Let's use moco only for TensorFlow Frontend modules.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
4 years ago[tf2tflite] Set shape information in ModelSignature (#8290)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 18 Oct 2019 00:42:46 +0000 (09:42 +0900)]
[tf2tflite] Set shape information in ModelSignature (#8290)

This commit will enable setting shape information in `ModelSignature`
at `tf2tflite`

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
4 years ago[nnkit-mocotf] Set shape information in ModelSignature (#8289)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 18 Oct 2019 00:42:38 +0000 (09:42 +0900)]
[nnkit-mocotf] Set shape information in ModelSignature (#8289)

This commit will enable setting shape information in `ModelSignature`
at `nnkit-mocotf`

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
4 years ago[res] Tensorflow test file for Relu( Add( Conv2D, Scalar )) (#8282)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 23:33:31 +0000 (08:33 +0900)]
[res] Tensorflow test file for Relu( Add( Conv2D, Scalar )) (#8282)

This adds Conv2D + Add + ReLU where Add's Const operand is a scalar (size 1 tensor).
To check exo handles this OK.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[moco-integration-test] Enable MergeConcatPass tests (#8288)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 23:19:27 +0000 (08:19 +0900)]
[moco-integration-test] Enable MergeConcatPass tests (#8288)

This commit will enable `MergeConcatPass` related tests
in `moco-integration-test`

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
4 years ago[tf2tflite] Enable MergeConcatPass tests (#8287)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 22:16:06 +0000 (07:16 +0900)]
[tf2tflite] Enable MergeConcatPass tests (#8287)

This commit will enable `MergeConcatPass` related tests

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
4 years ago[exo] Enable ConvertTensorConcat knob (#8286)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 22:15:42 +0000 (07:15 +0900)]
[exo] Enable ConvertTensorConcat knob (#8286)

This commit will enable `ConvertTensorConcat` knob in `exo`

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
4 years ago[ncc/interpreter] Fix Reshape operation for quantization (#8246)
Pavel Iliutchenko/AI Tools Lab /SRR/Engineer/Samsung Electronics [Thu, 17 Oct 2019 18:45:38 +0000 (21:45 +0300)]
[ncc/interpreter] Fix Reshape operation for quantization (#8246)

* Made Reshape independent from DataType
* Fixed setting output quantization

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
4 years ago[nnc] Fix constant folding optimization for quantization (#8241)
Pavel Iliutchenko/AI Tools Lab /SRR/Engineer/Samsung Electronics [Thu, 17 Oct 2019 18:43:36 +0000 (21:43 +0300)]
[nnc] Fix constant folding optimization for quantization (#8241)

* Fixed transpose for different DataType
* Added qunatization setting on optimization

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
4 years ago[nnc] Fix ModelAnalyzer.linearization test (#8291)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Thu, 17 Oct 2019 15:22:00 +0000 (18:22 +0300)]
[nnc] Fix ModelAnalyzer.linearization test (#8291)

The test relied on specific order of graph traversal.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[moco] Update input accessor (#8279)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 10:15:42 +0000 (19:15 +0900)]
[moco] Update input accessor (#8279)

This will update to use correct input accessor for MaxPool and FusedBatchNorm node in moco
and remove temporary added accessors

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[exo] Add tests for TFLConcatenate (#8273)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 10:15:17 +0000 (19:15 +0900)]
[exo] Add tests for TFLConcatenate (#8273)

This commit will add tests for `TFLConcatenate` in `exo`

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
4 years ago[moco-tf] Introduce TFNodeBuildTester and apply BiasAdd test (#8272)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 10:14:33 +0000 (19:14 +0900)]
[moco-tf] Introduce TFNodeBuildTester and apply BiasAdd test (#8272)

This will introduce TFNodeBuildTester test helper class for testing node's GraphBuilder class.
And applied BiasAdd test code is also shown to show how it works.

This change purpose is not to use Import class and to test only the GraphBuilder that
converts TensorFlow NodeDef to TF dialect node.

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[exo] Introduce MergeConcatNodePass (#8264)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 10:14:15 +0000 (19:14 +0900)]
[exo] Introduce MergeConcatNodePass (#8264)

* [exo] Introduce MergeConcatNodePass

This commit will introduce `MergeConcatNodePass`,
which fold multiple `TFLConcatenation` nodes with same axis

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
* apply comments

* add FusedActFunc check

4 years ago[locomotiv] Introduce TensorConstantPad (#8027)
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 10:04:16 +0000 (19:04 +0900)]
[locomotiv] Introduce TensorConstantPad (#8027)

* [locomotiv] Introduce TensorConstantPad

This commit introduces TensorConstantPad to locomotiv.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>
* change algorithm of pad interpretation.

* make constant value more general.

* change assert to validate.

* modify test.

* add assert for pad_data.

* constant input can be all operator.

* fix some wrong valiate and message.

4 years ago[moco-tf] Remove Shape Type inference header (#8270)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 09:58:56 +0000 (18:58 +0900)]
[moco-tf] Remove Shape Type inference header (#8270)

This will remove deprecated Shape and Type inference header use that of moco

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[moco-tf] Update input accessor (#8276)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 09:40:50 +0000 (18:40 +0900)]
[moco-tf] Update input accessor (#8276)

This will update to use correct input accessor for MaxPool and FusedBatchNorm node and related variables

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[tool] Support SquaredDifferene in select operator (#8263)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 17 Oct 2019 08:27:09 +0000 (17:27 +0900)]
[tool] Support SquaredDifferene in select operator (#8263)

- Support SquaredDifferene in select operator
- Add comment: not supporting options on schema v1.13.1

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years ago[exo] Register TransposedConv2DConverter (#8271)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 07:38:31 +0000 (16:38 +0900)]
[exo] Register TransposedConv2DConverter (#8271)

This commit registers TransposedConv2DConverter to convert stage.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[exo] adds loop to fuse the nodes for FuseReluPass (#8266)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 07:23:06 +0000 (16:23 +0900)]
[exo] adds loop to fuse the nodes for FuseReluPass (#8266)

This completes FuseReluPass.cpp by add loops that performs fusing selected nodes.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[plier-tf] Introduce parse_nodedef (#8261)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 06:49:36 +0000 (15:49 +0900)]
[plier-tf] Introduce parse_nodedef (#8261)

This will introduce parse_nodedef() method to parse tensorflow::NodeDef object

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[exo] Introduce TFLConcatenation exporter (#8259)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 06:49:26 +0000 (15:49 +0900)]
[exo] Introduce TFLConcatenation exporter (#8259)

This commit will introduce `TFLConcatenation` exporter in `exo`

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
4 years ago[exo] Convert canonical TransposedConv2D (#8267)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 06:35:25 +0000 (15:35 +0900)]
[exo] Convert canonical TransposedConv2D (#8267)

This commit introduces convert stage for canonical TransposedConv2D.
This conversion would apply to TransposedConv2D to make TFL-dialect
TFLTransposeConv with additional auxiliary nodes.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[exo] Export TFLTransposeConv (#8260)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 06:26:13 +0000 (15:26 +0900)]
[exo] Export TFLTransposeConv (#8260)

This commit introduces export stage of TFLTransposeConv operator

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[moco-tf] Redirect to moco Service (#8265)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 06:25:28 +0000 (15:25 +0900)]
[moco-tf] Redirect to moco Service (#8265)

This will update Shape and Type inference class to be redirected to moco Service and remove unused source files

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[exo] TransposeConv type & shape inference (#8258)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 05:28:38 +0000 (14:28 +0900)]
[exo] TransposeConv type & shape inference (#8258)

This commit introduces type and shape inferences of TFLTransposeConv

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[nnpkg-spec] Add instance_norm and revise structure and level (#8148)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 04:41:16 +0000 (13:41 +0900)]
[nnpkg-spec] Add instance_norm and revise structure and level (#8148)

It updates nnpackage specification to define `instance_norm` operator.
Also, it updates the document structure since it begin to have additional
operator. I move operators to same level of model for user to find them easily.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
4 years ago[exo] adding TFLRelu6Converter into convert.cpp (#8253)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 04:28:44 +0000 (13:28 +0900)]
[exo] adding TFLRelu6Converter into convert.cpp (#8253)

This adds TFLRelu6Converter into convert.cpp with related knob on.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[exo] TransposeConv IR (#8237)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 04:18:11 +0000 (13:18 +0900)]
[exo] TransposeConv IR (#8237)

* [exo] TransposeConv IR

This commit introduces TFLTransposeConv IR, TFL-dialect of loco node
representing TFlite TransposedConv operator.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Use camel case member name

4 years ago[res] Add simple network of placeholder with unknown dimension. (#8229)
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 17 Oct 2019 03:19:23 +0000 (12:19 +0900)]
[res] Add simple network of placeholder with unknown dimension. (#8229)

It is a simple network of placeholder with unknown dimension.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>
4 years ago[exo] refactoring using new function (`create_fused_bias()`) (#8230)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 03:18:38 +0000 (12:18 +0900)]
[exo] refactoring using new function (`create_fused_bias()`) (#8230)

This replaces body with a statement that calls extracted function. (`create_fused_bias()`)

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[lib/tflite] Update nnapi delegate (#8238)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 17 Oct 2019 02:15:45 +0000 (11:15 +0900)]
[lib/tflite] Update nnapi delegate (#8238)

Update nnapi delegate to match with v1.13.1

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years agoRename srcn backend shape (#8120)
여지환/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 17 Oct 2019 01:50:06 +0000 (10:50 +0900)]
Rename srcn backend shape (#8120)

Rename srcn backend Shape class into TensorDescriptor
      - Use same class name as cpu backend
      - Rename variables according to new class name

Signed-off-by: JiHwan Yeo <jihwan.yeo@samsung.com>
4 years ago[moco] Introduce Shape and Type inference Service (#8228)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 17 Oct 2019 01:07:01 +0000 (10:07 +0900)]
[moco] Introduce Shape and Type inference Service (#8228)

* [moco] Introduce Shape and Type inference Service

This will introduce Shape and Type inference Service from copied from moco-tf

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* make compiler happy

* another for compiler

4 years ago[neurun] Introduce build option for default config (#8167)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 17 Oct 2019 00:22:20 +0000 (09:22 +0900)]
[neurun] Introduce build option for default config (#8167)

- Introduce ENVVAR_NEURUN_CONFIG option: use default neurun config using environment (default: on)
- Default off on tizen, and introduce gbs build option "test_build" to enable environment variable setting

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years ago[nnc] Explicitly declare some variables as `Operation::Input *` (#8247)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 19:05:41 +0000 (22:05 +0300)]
[nnc] Explicitly declare some variables as `Operation::Input *` (#8247)

This is needed for transition period.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[mir-tflite-importer] Fixed a bug (wrong Deconv2d constructir). (#8248)
Gusev Dmitry/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 19:00:12 +0000 (22:00 +0300)]
[mir-tflite-importer] Fixed a bug (wrong Deconv2d constructir). (#8248)

Setting attributes values is corrected.

Signed-off-by: Dmitry Gusev <d.gusev@partner.samsung.com>
4 years ago[mir-caffe-importer] Fixed a bug in CaffeOpCreator (#8243)
Gusev Dmitry/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 18:43:55 +0000 (21:43 +0300)]
[mir-caffe-importer] Fixed a bug in CaffeOpCreator (#8243)

Setting attributes values is corrected.

Signed-off-by: Dmitry Gusev <d.gusev@partner.samsung.com>
4 years ago[nnc/interpreter] Fix Transpose for quantization (#8242)
Pavel Iliutchenko/AI Tools Lab /SRR/Engineer/Samsung Electronics [Wed, 16 Oct 2019 18:31:57 +0000 (21:31 +0300)]
[nnc/interpreter] Fix Transpose for quantization (#8242)

* Fixed quantization for output
* Fixed work with every DataType

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
4 years ago[mir2loco] Replace getInputProducer with getInput (#8245)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 18:29:45 +0000 (21:29 +0300)]
[mir2loco] Replace getInputProducer with getInput (#8245)

Replace calls to `getInputProducer` with equivalent `getInput`.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[nnc] Replace getInputProducer with getInput (#8244)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 18:28:52 +0000 (21:28 +0300)]
[nnc] Replace getInputProducer with getInput (#8244)

Replace calls to `getInputProducer` with equivalent `getInput`.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[mir] Make getInput the same as getInputProducer (#8236)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 13:28:25 +0000 (16:28 +0300)]
[mir] Make getInput the same as getInputProducer (#8236)

Make `getInput` the same as `getInputProducer`. The latter will be removed.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[mir-onnx-importer] Use new operations interface in importer. (#8182)
Gusev Dmitry/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 10:56:40 +0000 (13:56 +0300)]
[mir-onnx-importer] Use new operations interface in importer.  (#8182)

* Using new operations interface in importer.

Changes due to new operations constructors signatures.

Signed-off-by: Dmitry Gusev <d.gusev@partner.samsung.com>
4 years ago[mir2loco] Switch to `getInputProducer` (#8190)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 10:55:04 +0000 (13:55 +0300)]
[mir2loco] Switch to `getInputProducer` (#8190)

Replace calls to `getInput()->getProducer()` with calls to `getInputProducer()`.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[nnc] Switch to `getInputProducer` method (#8189)
Sergei Barannikov/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 16 Oct 2019 10:53:24 +0000 (13:53 +0300)]
[nnc] Switch to `getInputProducer` method (#8189)

Replace calls to `getInput()->getProducer()` with calls to `getInputProducer()`.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
4 years ago[angkor] Overload operator== (#8233)
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 16 Oct 2019 10:40:57 +0000 (19:40 +0900)]
[angkor] Overload operator== (#8233)

* [angkor] Overload operator==

This commit overloads operator== to index class.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>
* apply commnet.

4 years ago[angkor] Overload operator+ (#8210)
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 16 Oct 2019 09:50:16 +0000 (18:50 +0900)]
[angkor] Overload operator+ (#8210)

* [angkor] Overload operator+

This commit overloads operator+ to Index class.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>
* add out_of_range thorw exception.

* remove unnecessary header.

* change error type.

* change operator+ from member function to non-member.

4 years agolocomotiv is static library (#8212)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 16 Oct 2019 09:09:17 +0000 (18:09 +0900)]
locomotiv is static library (#8212)

This commit changes locomotiv from shared to static library.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[moco-tf] Remove VariadicArityNode (#8227)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 16 Oct 2019 08:27:19 +0000 (17:27 +0900)]
[moco-tf] Remove VariadicArityNode (#8227)

This will remove not used anymore VariadicArityNode and its test

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[moco-tf] Remove IR tests (#8225)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 16 Oct 2019 08:23:47 +0000 (17:23 +0900)]
[moco-tf] Remove IR tests (#8225)

This will remove IR test codes that already moved to moco-lang

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[exo] Introduce TFLConcatenation converter (#8221)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 16 Oct 2019 08:03:58 +0000 (17:03 +0900)]
[exo] Introduce TFLConcatenation converter (#8221)

This commit will introduce `TFLConcatenation` converter in `exo`

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
4 years ago[exo] Refactoring: Extract code from big method (Fuse) (#8179)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 16 Oct 2019 08:03:47 +0000 (17:03 +0900)]
[exo] Refactoring: Extract code from big method (Fuse) (#8179)

A piece of code is extracted as a method (`create_fused_bias()`) from big method (`Fuse`).

For diff difficulty, old code will be replace with one single call in next PR.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[exo] TFLMixin for classes that have fused activation function (#8223)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 16 Oct 2019 07:53:35 +0000 (16:53 +0900)]
[exo] TFLMixin for classes that have fused activation function (#8223)

This adds TFLMixin class for TFL nodes that have fused activation function

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
4 years ago[nnfw-api] use circle loader for circle model (#8222)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 16 Oct 2019 07:51:49 +0000 (16:51 +0900)]
[nnfw-api] use circle loader for circle model (#8222)

It uses model_types from MANIFEST to determine loader.
It also added missing namespace neurun for circle_loader.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
4 years ago[nnpkg-test] Remove conv test (#8220)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 16 Oct 2019 07:51:33 +0000 (16:51 +0900)]
[nnpkg-test] Remove conv test (#8220)

Remove fail conv test: neurun cannot run convolution with input weight

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
4 years ago[exo] TODO rigorous padding conversion (#8219)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 16 Oct 2019 07:48:34 +0000 (16:48 +0900)]
[exo] TODO rigorous padding conversion (#8219)

This commit adds TODO comments for rigorous padding conversion

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
4 years ago[exo] Introducing FuseReluPass (#8176)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 16 Oct 2019 07:48:10 +0000 (16:48 +0900)]
[exo] Introducing FuseReluPass (#8176)

* [exo] Introducing FuseReluPass

This introduces FuseReluPass, which fuse relu/relu6 into 10 TensorFlow Lite ops.
This commit adds code that finds candidate subgraph.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* remove unused var (to avoid CI error)

* Update compiler/exo/src/Pass/FuseReluPass.h

Co-Authored-By: hyunsik-yoon <hyunsik.yoon@samsung.com>
4 years ago[moco-tf] Revise IR to using moco::IR (#8218)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 16 Oct 2019 07:39:12 +0000 (16:39 +0900)]
[moco-tf] Revise IR to using moco::IR (#8218)

This will revise moco::tf IR and base classes to alias of moco IR and base classes

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[exo] Register FoldTransposeOfConst pass (#8202)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 16 Oct 2019 06:59:54 +0000 (15:59 +0900)]
[exo] Register FoldTransposeOfConst pass (#8202)

This commit registers FoldTransposeOfConst pass to exo TFL-level
optimization.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>