platform/core/ml/nnfw.git
5 years ago[neurun] Replace OperandIndex in LocalResponseNormalization::Param (#8708)
Sergei Barannikov/AI Tools Lab /SRR/Engineer/Samsung Electronics [Mon, 4 Nov 2019 02:27:59 +0000 (05:27 +0300)]
[neurun] Replace OperandIndex in LocalResponseNormalization::Param (#8708)

Replace `OperandIndex` in `LocalResponseNormalization::Param` with `int32_t` / `float`.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[lib/ncnn] Fix depthwise conv NCHW kernel bug (#8711)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 4 Nov 2019 02:27:36 +0000 (11:27 +0900)]
[lib/ncnn] Fix depthwise conv NCHW kernel bug (#8711)

Fix depthwise convolution NCHW 3x3 weight same padding kernel bug

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[base_loader] Throw when file open failed (#8706)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 4 Nov 2019 01:09:26 +0000 (10:09 +0900)]
[base_loader] Throw when file open failed (#8706)

Throw with a brief message when file open failed

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[exo] TFLMean IR (#8698)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Sun, 3 Nov 2019 22:09:46 +0000 (07:09 +0900)]
[exo] TFLMean IR (#8698)

This commit introduces TFLMean IR

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco] Fix typo on TFMean shape inference (#8699)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 08:24:27 +0000 (17:24 +0900)]
[moco] Fix typo on TFMean shape inference (#8699)

This commit fixes typo on TFMean shape inference

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[nnpackage_run] Use HDF5 C++ interface (#8671)
Sergei Barannikov/AI Tools Lab /SRR/Engineer/Samsung Electronics [Fri, 1 Nov 2019 08:19:32 +0000 (11:19 +0300)]
[nnpackage_run] Use HDF5 C++ interface (#8671)

Use HDF5 C++ interface instead of C.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoDo NOT install locomotiv (#8690)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 1 Nov 2019 07:46:24 +0000 (16:46 +0900)]
Do NOT install locomotiv (#8690)

locomotiv is now built as a static library, and thus it is no longer
necessary to install this library.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco] Use moco Importer.h (#8702)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 07:35:39 +0000 (16:35 +0900)]
[moco] Use moco Importer.h (#8702)

This will update to use moco/Importer.h in moco and remove old file in sub folder

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[test] Fix misleading path (#8693)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 07:29:52 +0000 (16:29 +0900)]
[test] Fix misleading path (#8693)

Nit: fix path in `test_arm_neurun_acl_cl.sh`.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[neurun] Introduce InstanceNorm op (#8526)
여지환/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 1 Nov 2019 07:23:55 +0000 (16:23 +0900)]
[neurun] Introduce InstanceNorm op (#8526)

This commit introduces InstanceNorm op into runtime frontend.
  - Introduce InstanceNormNode into neurun
  - Introduce InstanceNorm op into frontend
  - Use default value if value of epsilon is zero

Signed-off-by: JiHwan Yeo <jihwan.yeo@samsung.com>
5 years ago[exo] Knob for fuse SquaredDifference (#8696)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 07:15:05 +0000 (16:15 +0900)]
[exo] Knob for fuse SquaredDifference (#8696)

This commit introduces UseFuseSquaredDifferencePass knob

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[exo] Register EltwiseSqrt Converter (#8697)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 07:14:51 +0000 (16:14 +0900)]
[exo] Register EltwiseSqrt Converter (#8697)

This commit registers EltwiseSqrt Converter to exo conversion stage

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[exo] Remove redundant steps of type and shape inference (#8695)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 07:14:01 +0000 (16:14 +0900)]
[exo] Remove redundant steps of type and shape inference (#8695)

This removes redundant steps of type and shape inference.

Note that convert() and optimize() has already performed type and shape inference.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[angkor] Change num_elements behavior (#8684)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 1 Nov 2019 05:44:48 +0000 (14:44 +0900)]
[angkor] Change num_elements behavior (#8684)

The current implementation treats rank-0 tensors as a special tensor
which has no elements.

This semantics has a conflict with the well-known standard which treats
a rank-0 tensor as a scalar.

This commit resolves this semantic conflict.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[exo] Fuse Squared Difference (#8654)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 05:36:01 +0000 (14:36 +0900)]
[exo] Fuse Squared Difference (#8654)

This commit introduces a pass to fuse Squared Difference pattern

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Remove alias and use moco (#8688)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 05:20:59 +0000 (14:20 +0900)]
[moco-tf] Remove alias and use moco (#8688)

This will remove moco::tf alias for moco and update to use moco types

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years agoIntroduce Readers for nchw (#8663)
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 04:17:23 +0000 (13:17 +0900)]
Introduce Readers for nchw (#8663)

This commit introduces Readers for nchw.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years ago[exo] Export Sqrt, Rsqrt, SquaredDifference (#8652)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 04:13:12 +0000 (13:13 +0900)]
[exo] Export Sqrt, Rsqrt, SquaredDifference (#8652)

This commit implements export stage for TFLSqrt, TFLRsqrt and
TFLSquaredDifference. Note that both TFLite exporter and Circle exporter
implemented.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Use moco/Importer.h (#8686)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 04:10:45 +0000 (13:10 +0900)]
[moco-tf] Use moco/Importer.h (#8686)

This will update to use moco/Importer.h in moco-tf

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[nnkit-mocotf] Use moco namespace types not alias (#8682)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 03:29:38 +0000 (12:29 +0900)]
[nnkit-mocotf] Use moco namespace types not alias (#8682)

This will update to not to use alias moco::tf namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[tf2circle] Use moco namespace types not alias (#8681)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 03:29:03 +0000 (12:29 +0900)]
[tf2circle] Use moco namespace types not alias (#8681)

This will update to not to use alias moco::tf namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[nnkit-tf] Throw T/F error message (#8673)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 1 Nov 2019 03:03:38 +0000 (12:03 +0900)]
[nnkit-tf] Throw T/F error message (#8673)

The exception thrown inside run method now contains detailed message
that comes from TensorFlow itself.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[neurun] Update ifdef guards of operations (#8683)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 03:01:37 +0000 (12:01 +0900)]
[neurun] Update ifdef guards of operations (#8683)

Update `#ifdef` guards of operations which were not updated when it was
renamed.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[nnapi] Remove skipping Gather operation on cpu tests (#8667)
Pavel Iliutchenko/AI Tools Lab /SRR/Engineer/Samsung Electronics [Fri, 1 Nov 2019 02:18:28 +0000 (05:18 +0300)]
[nnapi] Remove skipping Gather operation on cpu tests (#8667)

* Enabled tests for Gather operation on cpu backends

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
5 years ago[tf2tflite] Use moco namespace types not alias (#8680)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 01:54:20 +0000 (10:54 +0900)]
[tf2tflite] Use moco namespace types not alias (#8680)

This will update to not to use alias moco::tf namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco] Tidy import unused dependency (#8676)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 01:49:07 +0000 (10:49 +0900)]
[moco] Tidy import unused dependency (#8676)

This will remove unused dependency headers and libraries

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[exo] Fuse Rsqrt (#8655)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 01:40:07 +0000 (10:40 +0900)]
[exo] Fuse Rsqrt (#8655)

This commit introduces a pass to fuse Rsqrt pattern

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[exo] Convert EltwiseSqrt (#8653)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 01:19:19 +0000 (10:19 +0900)]
[exo] Convert EltwiseSqrt (#8653)

This commit implements conversion from canonical EltwiseSqrt to
TFL-dialect TFLSqrt

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco] Provide Importer header in moco folder (#8672)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 1 Nov 2019 01:08:30 +0000 (10:08 +0900)]
[moco] Provide Importer header in moco folder (#8672)

This will provide Importer header in moco folder

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago [neurun] Rename operation classes (#8666)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 1 Nov 2019 00:58:26 +0000 (09:58 +0900)]
 [neurun] Rename operation classes (#8666)

This commit removes suffix 'Node' from operation class names since
redundant.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[tflite_loader] Introduce tflite_loader test list (#8660)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 31 Oct 2019 12:09:13 +0000 (21:09 +0900)]
[tflite_loader] Introduce tflite_loader test list (#8660)

Itnroduce tflite_loader test list for tflite_loader validation test

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[runtime] Move runtimes and rename runtime directory (#8472)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 31 Oct 2019 10:30:06 +0000 (19:30 +0900)]
[runtime] Move runtimes and rename runtime directory (#8472)

Move logging and pacl runtime into contrib
Rename runtime directory: runtimes -> runtime

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[moco] Update Importer comment (#8657)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 10:04:52 +0000 (19:04 +0900)]
[moco] Update Importer comment (#8657)

This will update Importer comment with input and output nodes

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Remove output node handle for Push (#8656)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 10:04:30 +0000 (19:04 +0900)]
[moco-tf] Remove output node handle for Push (#8656)

This will remove not needed anymore handle for Canonical Push as output node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[base_loader] Remove unused variable (#8659)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 31 Oct 2019 09:57:59 +0000 (18:57 +0900)]
[base_loader] Remove unused variable (#8659)

Remove unused variable in `BaseLoader::loadOperand`.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[neurun] Inline Graph::linearize (#8627)
Sergei Barannikov/AI Tools Lab /SRR/Engineer/Samsung Electronics [Thu, 31 Oct 2019 09:31:29 +0000 (12:31 +0300)]
[neurun] Inline Graph::linearize (#8627)

This is to remove dependency of `Graph` on `Linear`.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoMake Readers and Views support both ITensor and buffer (#8644)
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 31 Oct 2019 09:30:56 +0000 (18:30 +0900)]
Make Readers and Views support both ITensor and buffer (#8644)

This commit makes Readers and Views support both ITensor and buffer.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years ago[neurun] Allow same inputs multiple times (#8567)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 31 Oct 2019 09:15:08 +0000 (18:15 +0900)]
[neurun] Allow same inputs multiple times (#8567)

Allow same inputs multiple times for an operation.
Remove redundant assertion

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[neurun] Support 1x1x1x1 input for Squeeze (#8646)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 09:11:53 +0000 (18:11 +0900)]
[neurun] Support 1x1x1x1 input for Squeeze (#8646)

Squeeze will handle 1x1x1x1 input using ShapeFixer.
Scalar output will be fixed to rank-1 one element tensor.
The output value will be transfer to app without trouble since nnfw
runtime API get output values using buffer pointer and buffer length.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
5 years ago[moco] Produce TFPush for graph output (#8647)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 07:47:27 +0000 (16:47 +0900)]
[moco] Produce TFPush for graph output (#8647)

* [moco] Produce TFPush for graph output

This will update to produce TFPush for graph output node

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

5 years ago[exo] Type & shape infer. for Sqrt, Rsqrt, SquaredDifference (#8649)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 31 Oct 2019 07:19:45 +0000 (16:19 +0900)]
[exo] Type & shape infer. for Sqrt, Rsqrt, SquaredDifference (#8649)

This commit introduces type and shape inferences or TFLSqrt, TFLRsqrt,
TFLSquaredDifference.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[circle-verify] Add VerifyFlatBuffers (#8648)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 06:40:49 +0000 (15:40 +0900)]
[circle-verify] Add VerifyFlatBuffers (#8648)

This will add and enable VerifyFlatBuffers verify Circle file in flatbuffer file format level

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[exo] Note on Pass name convention (#8645)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 31 Oct 2019 06:37:52 +0000 (15:37 +0900)]
[exo] Note on Pass name convention (#8645)

This commit adds note on Pass name convention

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[exo] Rename to have Pass suffix (#8643)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 31 Oct 2019 05:19:19 +0000 (14:19 +0900)]
[exo] Rename to have Pass suffix (#8643)

This commit renames following classes (and related files) to have 'Pass'
suffix like other Passes:
- FoldReshapeOfConst -> FoldReshapeOfConstPass
- FoldTransposeOfConst -> FoldTransposeOfConstPass

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[exo] TFL nodes for Sqrt, Rsqrt, SquaredDifference (#8623)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 31 Oct 2019 05:14:49 +0000 (14:14 +0900)]
[exo] TFL nodes for Sqrt, Rsqrt, SquaredDifference (#8623)

This commit introduces simple TFL-dialect nodes:
- locoex::TFLSqrt
- locoex::TFLRsqrt
- locoex::TFLSquaredDifference

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[circle-verify] Introduce Model and empty Driver (#8641)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 05:07:27 +0000 (14:07 +0900)]
[circle-verify] Introduce Model and empty Driver (#8641)

* [circle-verify] Introduce Model and empty Driver

This will introdue Model that will provide mmapped buffer and empty Driver

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

5 years ago[moco-tf] Prepare to handle TFPush instead of Push (#8636)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 04:43:20 +0000 (13:43 +0900)]
[moco-tf] Prepare to handle TFPush instead of Push (#8636)

This will update Frontend to support output as TFPush and Push while updating to produce TFPush as Output

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[neurun] Remove workaround after switching to ACL v19.05 (#8629)
Sergei Barannikov/AI Tools Lab /SRR/Engineer/Samsung Electronics [Thu, 31 Oct 2019 04:34:24 +0000 (07:34 +0300)]
[neurun] Remove workaround after switching to ACL v19.05 (#8629)

Prior versions had a bug which did not take into accout the activation fused into CLDepthwiseConvolutionLayer. The workaround for this is now removed.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[nnpkg_run] Fix hdf5 file dump (#8575)
Vladimir Plazun/Engineer/AI Tools Lab /SRR/Samsung Electronics [Thu, 31 Oct 2019 04:24:29 +0000 (07:24 +0300)]
[nnpkg_run] Fix hdf5 file dump (#8575)

Value group have to be created only once for file
Remove blank line

Signed-off-by: Vladimir Plazun <v.plazun@samsung.com>
5 years agoChange the dimension order of Reader and Views in nhwc (#8509)
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 31 Oct 2019 04:15:17 +0000 (13:15 +0900)]
Change the dimension order of Reader and Views in nhwc (#8509)

This commit changes dimension order of Reader and Views in nhwc to nhwc.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years ago[circle-verify] Readme (#8635)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 01:56:01 +0000 (10:56 +0900)]
[circle-verify] Readme (#8635)

* [circle-verify] Readme

This will introduce circle-verify readme file explaining it's usage

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* edit about

* change output format

5 years ago[moco-tf] Add TFPush NodeSummaryBuilder (#8634)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 01:55:25 +0000 (10:55 +0900)]
[moco-tf] Add TFPush NodeSummaryBuilder (#8634)

This will add TFPush node to NodeSummmaryBuilder

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-integration-test] adding tests related to depthwiseConv2D bias fusion (#8543)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 01:47:44 +0000 (10:47 +0900)]
[moco-integration-test] adding tests related to depthwiseConv2D bias fusion (#8543)

This adds tests related to depthwiseConv2D bias fusion.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[tfl-inspect] Enable nullptr for data in buffer_info (#8632)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 01:34:15 +0000 (10:34 +0900)]
[tfl-inspect] Enable nullptr for data in buffer_info (#8632)

This will enable buffer_info() to accept nullptr in buffer_data for use cases where only return size is needed

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco] TF dialect OutputIndexQueryService (#8630)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 01:31:52 +0000 (10:31 +0900)]
[moco] TF dialect OutputIndexQueryService (#8630)

This will enable TF dialect GraphOutputIndexQueryService

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[tfl-inspect] Enable conv2d_weight option (#8633)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 01:29:03 +0000 (10:29 +0900)]
[tfl-inspect] Enable conv2d_weight option (#8633)

This will enable conv2d_weight option that shows Conv2D series node with non Const weight input

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Enable TFPush canonicalizer (#8631)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 01:27:08 +0000 (10:27 +0900)]
[moco-tf] Enable TFPush canonicalizer (#8631)

* [moco-tf] Enable TFPush canonicalizer

This will enable TFPushCanonicalizer in Canonicalizer phase

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix typo

5 years ago[neurun] fix loadSqueeze and SqueezeNode. (#8602)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 31 Oct 2019 01:03:48 +0000 (10:03 +0900)]
[neurun] fix loadSqueeze and SqueezeNode. (#8602)

- `loadSqueeze` should read `squeeze_dims` from options.
- `squeeze_dims` should be array in SqueezeNode.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
5 years ago[exo] Clean empty line for conversion registry (#8622)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 30 Oct 2019 23:30:41 +0000 (08:30 +0900)]
[exo] Clean empty line for conversion registry (#8622)

This commit clears empty line in Convert.h, which registers all
conversion passes. This is mainly to enhance readability.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[compiler] introducing tf2tflite-quality-test (#8614)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 22:01:08 +0000 (07:01 +0900)]
[compiler] introducing tf2tflite-quality-test (#8614)

This introduces a new project called `tf2tflite-quality-test`, whose previous name was `dredd`.
The basic structure of this is very similar with `tf2tflite-model-test`.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[exo] revising TestGraph.h (#8618)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 22:00:39 +0000 (07:00 +0900)]
[exo] revising TestGraph.h (#8618)

This revises TestGraph.h to 1) remove build() and 2) remove graph().
1) build() is moved into constructor
2) graph() is already defined in parent

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[moco-tf] Push canonicalizer (#8617)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 21:56:01 +0000 (06:56 +0900)]
[moco-tf] Push canonicalizer (#8617)

This will introduce Canonicalizer for TFPush node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[neurun] Support Gather operation on cpu backend (#8399)
Pavel Iliutchenko/AI Tools Lab /SRR/Engineer/Samsung Electronics [Wed, 30 Oct 2019 12:38:44 +0000 (15:38 +0300)]
[neurun] Support Gather operation on cpu backend (#8399)

* Added Gather kernel on cker
* Changed axis Gather param
* Fixed other parts with Gather

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
5 years ago[nnfw-gbs] nnfw-devel package config file (#8518)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 30 Oct 2019 10:11:17 +0000 (19:11 +0900)]
[nnfw-gbs] nnfw-devel package config file (#8518)

Introduce nnfw-dev package config file (nnfw.pc)
Change include path to {prefix}/include/nnfw

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[exo] Removing EnableTFLDialect knob (#8615)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 10:06:26 +0000 (19:06 +0900)]
[exo] Removing EnableTFLDialect knob (#8615)

This removes EnableTFLDialect knob - related files are modified.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[moco] push_node implementation (#8616)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 10:05:33 +0000 (19:05 +0900)]
[moco] push_node implementation (#8616)

This will introduce push_node implementation that retuns TFPush node with given GraphOutputIndex

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[tfl-inspect] Introduce DumpConv2DWeight (#8611)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 10:05:10 +0000 (19:05 +0900)]
[tfl-inspect] Introduce DumpConv2DWeight (#8611)

* [tfl-inspect] Introduce DumpConv2DWeight

This will introduce DumpConv2DWeight to dump Conv2D series node weight input type

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix namespace

* update comment

* use buildin_code

5 years ago[moco] Shape and type inference for TFPush (#8613)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 09:37:58 +0000 (18:37 +0900)]
[moco] Shape and type inference for TFPush (#8613)

This will add shape and type inference for TFPush node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco] Introduce TFPush IR (#8598)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 07:53:44 +0000 (16:53 +0900)]
[moco] Introduce TFPush IR (#8598)

This will introduce TFPush IR that will be used as graph output

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[lib/misc] Handle scalar input/output (#8607)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 30 Oct 2019 07:14:09 +0000 (16:14 +0900)]
[lib/misc] Handle scalar input/output (#8607)

Fix error when object is scalar
This will fix error nnapi_test when input/output is scalar

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[tfl-inspect] Readme section for operators option (#8605)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 06:51:00 +0000 (15:51 +0900)]
[tfl-inspect] Readme section for operators option (#8605)

This will introduce section in Readme file for operators option

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[tfl-inspect] Readme for conv2d_weight (#8606)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 06:50:36 +0000 (15:50 +0900)]
[tfl-inspect] Readme for conv2d_weight (#8606)

This will update Readme with conv2d_weight option

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[exo] Rename file name for type inference test (#8604)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 30 Oct 2019 06:32:17 +0000 (15:32 +0900)]
[exo] Rename file name for type inference test (#8604)

This commit renames following file:
TFLTypeInference.test.cpp -> TFLTypeInferenceRule.test.cpp

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[tf2tflite-model-test] Do NOT rename runner.sh (#8599)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 30 Oct 2019 06:27:08 +0000 (15:27 +0900)]
[tf2tflite-model-test] Do NOT rename runner.sh (#8599)

Now, copied script runner is named as "runner.sh" instead of "run".

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years agoIntroduce tf2circle conversion test (#8593)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 30 Oct 2019 06:26:54 +0000 (15:26 +0900)]
Introduce tf2circle conversion test (#8593)

This commit introduces tf2circle conversion test.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[neurun] initialize logging only once (#8570)
Vladimir Plazun/Engineer/AI Tools Lab /SRR/Samsung Electronics [Wed, 30 Oct 2019 03:38:40 +0000 (06:38 +0300)]
[neurun] initialize logging only once (#8570)

Currently logging::Context is instantiated inside each translation unit separately

Signed-off-by: Vladimir Plazun <v.plazun@samsung.com>
5 years ago[tfl-inspect] Add more examples with sort and uniq (#8579)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 01:48:35 +0000 (10:48 +0900)]
[tfl-inspect] Add more examples with sort and uniq (#8579)

This will add xample using pipe with sort and uniq

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Remove unused header (#8578)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 30 Oct 2019 01:33:06 +0000 (10:33 +0900)]
[moco-tf] Remove unused header (#8578)

This will remove unused header in TestHelper

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years agoIntroduce tf2circle model test (#8551)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 30 Oct 2019 00:13:53 +0000 (09:13 +0900)]
Introduce tf2circle model test (#8551)

This commit introduces tf2circle model test.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years agoRemove tf2tflite test (#8562)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 29 Oct 2019 23:45:56 +0000 (08:45 +0900)]
Remove tf2tflite test (#8562)

This commit removes tf2tflite_test in `compiler/tf2tflite` directory.
Note that this test is fully moved to `tf2tftlie-integration-test`.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Remove TFShapeInferenceHelper (#8576)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 29 Oct 2019 23:41:57 +0000 (08:41 +0900)]
[moco-tf] Remove TFShapeInferenceHelper (#8576)

This will remove not used anymore TFShapeInferenceHelper file that migrated to moco-support

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[mir-tflite-importer] Fix quantization creation with various data types (#8552)
Pavel Iliutchenko/AI Tools Lab /SRR/Engineer/Samsung Electronics [Tue, 29 Oct 2019 16:06:32 +0000 (19:06 +0300)]
[mir-tflite-importer] Fix quantization creation with various data types (#8552)

* Fixed creation quantization with empty quantization attribute on
tflite
* Removed data type checks

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
5 years ago[nnc] Remove file with obsolete documentation (#8574)
Alexander Efimov/./AI Tools Lab/Samsung Electronics [Tue, 29 Oct 2019 13:27:55 +0000 (16:27 +0300)]
[nnc] Remove file with obsolete documentation (#8574)

After nncc and nnfw repositories merge this file is not valid, because we use unified nnas format

Signed-off-by: Efimov Alexander <a.efimov@samsung.com>
5 years agoMove static vector `dimensions` inside `main()` (#8571)
Vladimir Plazun/Engineer/AI Tools Lab /SRR/Samsung Electronics [Tue, 29 Oct 2019 12:03:01 +0000 (15:03 +0300)]
Move static vector `dimensions` inside `main()` (#8571)

There is no need for it to have static lifetime

Signed-off-by: Vladimir Plazun <v.plazun@samsung.com>
5 years ago[ncnn-backend] Support add operation (#8565)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 29 Oct 2019 11:31:27 +0000 (20:31 +0900)]
[ncnn-backend] Support add operation (#8565)

- NCNN binary-op kernel porting (disable same input/output tensor support)
- NCNN backend add operation support
  - Support non-broadcasting
  - Need alignment matching when rank is upper 3
- Enable unittest, related validation test

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Fix subgraph creating for split (#8488)
Dilshodzhon Poshshoev/Engineer/AI Tools Lab /SRR/Samsung Electronics [Tue, 29 Oct 2019 11:24:51 +0000 (14:24 +0300)]
[neurun] Fix subgraph creating for split (#8488)

It merges split with one of the next operations. Because of that other
outputs of ths split aren't seen as outputs of a node and in LinearExecutor
topological sort doesn't work correct and it causes failure in memory
management because incorrect order of startLifetime/finishLifetime,
notifyFirstUse/notifyLastUse,...

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
5 years ago[nnfw-ncnn] Fix bug in Mat class (#8564)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 29 Oct 2019 08:42:53 +0000 (17:42 +0900)]
[nnfw-ncnn] Fix bug in Mat class (#8564)

Rename mat class code file mat.cpp to mat.cc - not compiled
Fix build error by call function in cpu.h

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[moco-tf] Use TFShapeInferenceHelper from moco another one (#8557)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 29 Oct 2019 07:56:04 +0000 (16:56 +0900)]
[moco-tf] Use TFShapeInferenceHelper from moco another one (#8557)

This will update to use TFShapeInferenceHelper from moco
- another update for Canonicalizers

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[nnpkg_test] Add missing option for nnpkg_test.sh (#8556)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 29 Oct 2019 07:44:37 +0000 (16:44 +0900)]
[nnpkg_test] Add missing option for nnpkg_test.sh (#8556)

This commit will fix `nnpkg_test.sh` to get proper options

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
5 years agoFix format script to use correct clang-format (#8558)
여지환/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 29 Oct 2019 04:04:03 +0000 (13:04 +0900)]
Fix format script to use correct clang-format (#8558)

Fix script to call correct version of clang-format that system supports

Signed-off-by: JiHwan Yeo <jihwan.yeo@samsung.com>
5 years ago[nnfw-ncnn] Restructure directory and introduce mat class (#8532)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 29 Oct 2019 03:51:21 +0000 (12:51 +0900)]
[nnfw-ncnn] Restructure directory and introduce mat class (#8532)

Restructure ncnn directory
- Introduce internal srcn directory to divide with original ncnn directory
- Introduce mat.cpp and mat.h
- Update copyright

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[exo] Remove operation conversion Knobs (#8517)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 29 Oct 2019 03:22:42 +0000 (12:22 +0900)]
[exo] Remove operation conversion Knobs (#8517)

* [exo] Remove operation conversion Knobs

This commit removes operation conversion Knobs. Available canonical
nodes would be converted to TFL dialect nodes with this commit by
default.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Remove unused, deprecated function

5 years ago[exo] Update Circle operation exporter (#8549)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 29 Oct 2019 03:16:50 +0000 (12:16 +0900)]
[exo] Update Circle operation exporter (#8549)

This commit updates Circle operation exporter to have same coverage of
TFlite operation exporter. Following operation export stages are
implemented:
- locoex::TFLConcatenation
- locoex::TFLDepthwiseConv2D
- locoex::TFLTransposeConv
- loco::TensorConstantPad

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years agoAdd Squeeze in base_loader (#8524)
이춘석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 29 Oct 2019 01:40:52 +0000 (10:40 +0900)]
Add Squeeze in base_loader (#8524)

- Add Squeeze in base_loader

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
5 years ago[android_jni] Add missing builtin_ops_jni.cc (#8548)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 29 Oct 2019 00:51:37 +0000 (09:51 +0900)]
[android_jni] Add missing builtin_ops_jni.cc (#8548)

This file should have been committed from #8026 but it was not added by
accident. Target `tensorflowlite_jni` would not build without this file.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoUpdate test project in collect resource script (#8546)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 29 Oct 2019 00:50:56 +0000 (09:50 +0900)]
Update test project in collect resource script (#8546)

`tf2tflite_test` would be deprecated, and replaced with
`tf2tflite_integration_test`. This commit updates regarding this change
for collect resource script.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Use TFShapeInferenceHelper from moco (#8545)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 29 Oct 2019 00:04:32 +0000 (09:04 +0900)]
[moco-tf] Use TFShapeInferenceHelper from moco (#8545)

This will update to use TFShapeInferenceHelper from moco

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[mir/interpreter] Quantized operations implementation (#8317)
Pavel Iliutchenko/AI Tools Lab /SRR/Engineer/Samsung Electronics [Mon, 28 Oct 2019 11:56:21 +0000 (14:56 +0300)]
[mir/interpreter] Quantized operations implementation (#8317)

* Supported quantized operations: Add, Conv2D, DepthwiseConv2D
* Fixed interpreter for quantizations

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
5 years ago[tfl-inspect] Remove unused header inc (#8544)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 10:08:26 +0000 (19:08 +0900)]
[tfl-inspect] Remove unused header inc (#8544)

This will remove unused header includes

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[tf2tflite-integration-test] adding test for DepthwiseConv2D bias fusion (#8542)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 09:47:43 +0000 (18:47 +0900)]
[tf2tflite-integration-test] adding test for DepthwiseConv2D bias fusion (#8542)

This adds four nets to test DepthwiseConv2D bias fusion.

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