platform/core/ml/nnfw.git
4 years ago[moco-tf] Replace assert in FuseBinaryIntoPreceding (#6009)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 30 Jul 2019 07:20:32 +0000 (16:20 +0900)]
[moco-tf] Replace assert in FuseBinaryIntoPreceding (#6009)

This will replace assert with a log for unknown as it is now know case and can be handled by other transforms

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[neurun] Remove hiding _indexed_ranks from Dataflow (#5926)
Дилшоджон Умронхонович Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Tue, 30 Jul 2019 06:46:01 +0000 (09:46 +0300)]
[neurun] Remove hiding _indexed_ranks from Dataflow (#5926)

_indexed_ranks is a protected member in both ExecutorBase and Dataflow.
ExecutorBase::setIndexedRanks() sets ExecutorBase::_indexed_ranks,
but methods in Dataflow uses Dataflow::_indexed_ranks and is always nullptr

Signed-off-by: Dilshodzhon Poshshoev <d.poshshoev@samsung.com>
4 years ago[moco-tf] Import as TFAvgPool with a knob (#6006)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 30 Jul 2019 06:18:07 +0000 (15:18 +0900)]
[moco-tf] Import as TFAvgPool with a knob (#6006)

This will add a Knob to import AvgPool node as TFAvgPool or AvgPool2D

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[moco/tf] DepthwiseConv2D FixShapeTransform bug fix. (#6003)
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 30 Jul 2019 04:52:46 +0000 (04:52 +0000)]
[moco/tf] DepthwiseConv2D FixShapeTransform bug fix. (#6003)

It needs to be modified for getting right stride data.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>
4 years ago[neurun] Enable {CL|NE}DepthwiseConvolutionLayer3x3 on KernelGenerator (#6008)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 30 Jul 2019 04:40:52 +0000 (13:40 +0900)]
[neurun] Enable {CL|NE}DepthwiseConvolutionLayer3x3 on KernelGenerator (#6008)

Enable {CL|NE}DepthwiseConvolutionLayer3x3 on each KernelGenerator. The
only CLDepthwiseConvolutionLayer3x3 has the parameter for the internal
buffer memory manager.

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
4 years ago[moco/tf] Add TFNodeSummaryBuilder for DepthwiseConv2dNative (#6004)
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 30 Jul 2019 04:33:59 +0000 (04:33 +0000)]
[moco/tf] Add TFNodeSummaryBuilder for DepthwiseConv2dNative (#6004)

This commit adds a TFNodeSummaryBuilder for DepthwiseConv2dNative.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>
4 years ago[neurun] Enable AclLinearMemoryManager on LinearExecutor (#5922)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 30 Jul 2019 04:07:41 +0000 (13:07 +0900)]
[neurun] Enable AclLinearMemoryManager on LinearExecutor (#5922)

Enable AclLinearMemoryManager to acl_cl and acl_neon  on LinearExecutor

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
4 years ago[moco-tf] Fix shape and pad for TFAvgPool (#5996)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 30 Jul 2019 03:43:35 +0000 (12:43 +0900)]
[moco-tf] Fix shape and pad for TFAvgPool (#5996)

This will implement FixShape and FixPadding for TFAvgPool node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[neurun] Fix preVisit/postVisit for double-concats (#5977)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 30 Jul 2019 03:36:57 +0000 (12:36 +0900)]
[neurun] Fix preVisit/postVisit for double-concats (#5977)

* [neurun] Fix preVisit/postVisit for double-concats

Fix preVisit and postVisit of TemplTensorBuilder to pass
double-concats(two consecutive concats) of
inception_v3_2018_04_27.tflite

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
* Fix comment

4 years agoMark NNPACK as NOT FOUND if PeachpySource is missing (#5995)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 30 Jul 2019 01:21:19 +0000 (10:21 +0900)]
Mark NNPACK as NOT FOUND if PeachpySource is missing (#5995)

The current implement says that NNPACK is available even though
PeachpySource is missing.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
4 years agoSet PeachpySource as NOT FOUND on failure (#5966)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 30 Jul 2019 00:38:01 +0000 (09:38 +0900)]
Set PeachpySource as NOT FOUND on failure (#5966)

The current implementation of PeachpySource package reports SUCCESS
even when source code generation fails.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
4 years ago[moco-tf] Revise importer test (#5960)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 30 Jul 2019 00:18:00 +0000 (09:18 +0900)]
[moco-tf] Revise importer test (#5960)

This will revise importer test to do test with TFIdentity IR that uses Identity node as a test material

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
4 years ago[moco-tf] Move vector<int64_t> streamout to LogHelper (#5981)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 29 Jul 2019 23:31:35 +0000 (08:31 +0900)]
[moco-tf] Move vector<int64_t> streamout to LogHelper (#5981)

This will move vector<int64_t> streamout of TFFormattedGraph to LogHelper

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Import Reshape operation as TF dialect (#5976)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 29 Jul 2019 22:24:54 +0000 (07:24 +0900)]
[moco-tf] Import Reshape operation as TF dialect (#5976)

This commit implements import stage of tensorflow Reshape operation to
TFReshape node.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Introduce as_tensor_shape with FeatureShape (#5973)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 29 Jul 2019 22:22:42 +0000 (07:22 +0900)]
[moco-tf] Introduce as_tensor_shape with FeatureShape (#5973)

This will introduce ShapeInferenceData helper to set as TensorShape with FeatureShape and TFDataLayout

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years agoFix #5978 (#5989)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 17:41:13 +0000 (20:41 +0300)]
Fix #5978 (#5989)

Adjust tests to take into account recent changes in compiler passes.
Fixes #5978.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[mir/Caffe importer] Introduce `importModel` method (#5986)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 17:37:38 +0000 (20:37 +0300)]
[mir/Caffe importer] Introduce `importModel` method (#5986)

Add `importModel` method as a future replacement for `import` and `createIR` methods.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[neurun] Add fine for scheduling into another backend (#5880)
Дилшоджон Умронхонович Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 12:54:11 +0000 (15:54 +0300)]
[neurun] Add fine for scheduling into another backend (#5880)

Assigning to another backend will cause creating a subgraph:
one for permutation and one more if it was mergable node
if the same backend was assigned

Signed-off-by: Dilshodzhon Poshshoev <d.poshshoev@samsung.com>
5 years ago[mir/TFLite importer] Introduce `importModel` method (#5988)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 12:49:06 +0000 (15:49 +0300)]
[mir/TFLite importer] Introduce `importModel` method (#5988)

Add `importModel` method as a future replacement for `import` and `createIR` methods.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[mir/ONNX importer] Introduce `importModel` method (#5987)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 12:48:45 +0000 (15:48 +0300)]
[mir/ONNX importer] Introduce `importModel` method (#5987)

Add `importModel` method as a future replacement for `import` and `createIR` methods.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[mir/Caffe2 importer] Introduce `importModel` method (#5985)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 12:48:17 +0000 (15:48 +0300)]
[mir/Caffe2 importer] Introduce `importModel` method (#5985)

Add `importModel` method as a future replacement for `import` and `createIR` methods.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[neurun] Add custom kernel registry (#5597)
Vladimir Plazun/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 12:43:24 +0000 (15:43 +0300)]
[neurun] Add custom kernel registry (#5597)

Add custom kernel registry used to keep track of custom kernels used by specific backend( see #5411 )

Signed-off-by: Vladimir Plazun <v.plazun@samsung.com>
5 years ago[nnc] Join `import` and `createIR` methods into `importModel` (#5984)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 11:35:55 +0000 (14:35 +0300)]
[nnc] Join `import` and `createIR` methods into `importModel` (#5984)

Methods `import` and `createIR` are always used together. Join them to simplify usage.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoFix unmatched result of tconv in NCHW (#5968)
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 29 Jul 2019 11:30:43 +0000 (20:30 +0900)]
Fix unmatched result of tconv in NCHW (#5968)

This commit fixes unmatched result of tconv in NCHW
  - Change weights of tconv to OIHW in case of NCHW

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years ago[moco-tf] Summary for TFReshape node (#5979)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 29 Jul 2019 11:16:12 +0000 (20:16 +0900)]
[moco-tf] Summary for TFReshape node (#5979)

This commit adds summary for TFReshape node

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Enable AvgPoolCanonicalizer (#5980)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 29 Jul 2019 11:15:43 +0000 (20:15 +0900)]
[moco-tf] Enable AvgPoolCanonicalizer (#5980)

This will add AvgPoolCanonicalizer to Canonicalizer list

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[loco] Introduce const Graph getters (#5974)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 29 Jul 2019 11:13:41 +0000 (20:13 +0900)]
[loco] Introduce const Graph getters (#5974)

Now, Graph provides const "nodes", "inputs", "outputs" methods.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[loco] Remove unused input_nodes helper (#5972)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 29 Jul 2019 11:13:31 +0000 (20:13 +0900)]
[loco] Remove unused input_nodes helper (#5972)

* [loco] Remove unused input_nodes helper

This commit removes unused input_nodes helper. This helper is currently
unused, but its presence makes it difficult to decouple loco core from
loco canonical.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Remove the implementation

5 years ago[loco] Input/Output Index from Push/Pull node (#5961)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 29 Jul 2019 11:13:19 +0000 (20:13 +0900)]
[loco] Input/Output Index from Push/Pull node (#5961)

* [loco] Input/Output Index from Push/Pull node

This commit revises Push/Pull node to store associated Graph-level
Input/Output index in it.

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

5 years ago[moco-tf] Introduce TFMaxPool IR (#5975)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 29 Jul 2019 06:21:05 +0000 (15:21 +0900)]
[moco-tf] Introduce TFMaxPool IR (#5975)

This will introduce TFMaxPool IR in TensorFlow Dialect for MaxPool node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[neurun] Always notify ExecutionObserver (#5958)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 29 Jul 2019 06:12:12 +0000 (15:12 +0900)]
[neurun] Always notify ExecutionObserver (#5958)

Now, DataflowExecutor notifies the beginning and end of the whole
execution whenever any observer exists (even when profiling mode is
disabled).

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years agoFix format error in moco-tf (#5970)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 29 Jul 2019 05:38:30 +0000 (14:38 +0900)]
Fix format error in moco-tf (#5970)

Fix format error in moco-tf's FixShapeTransform.cpp

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[moco/tf] Introduce TFDepthwiseConv2D IR. (#5912)
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 29 Jul 2019 05:12:18 +0000 (05:12 +0000)]
[moco/tf] Introduce TFDepthwiseConv2D IR. (#5912)

* [moco/tf] Introduce TFDepthwiseConv2D IR.

This commit introduce TFDepthwiseConv2D IR.

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

5 years agoRemove sinks and sources member in ExecutorBase (#5885)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 29 Jul 2019 04:25:16 +0000 (13:25 +0900)]
Remove sinks and sources member in ExecutorBase (#5885)

Remove sinks and sources member in ExecutorBase and source/sink methods return unique pointer
Move unused input checking into execute method
Remove unused parameter for source/sink

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Remove unnecessary class and verbose code (#5930)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 29 Jul 2019 01:49:15 +0000 (10:49 +0900)]
[neurun] Remove unnecessary class and verbose code (#5930)

Remove unnecessary class(LinearAclMemoryManager) and verbose code

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years agoUnified dockerfile (#5925)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 29 Jul 2019 01:42:29 +0000 (10:42 +0900)]
Unified dockerfile (#5925)

Merge infra/nncc/Dockerfile, infra/nnfw/docker/Dockerfile, and infra/nnfw/docker/Dockerfile_tizen into infra/docker/Dockerfile
- Base ubuntu version: 16.04
- Use apt-get
- Native/cross gcc/g++ version: 6.x
- Update nncc/nnfw docker build documents
- Move infra/nnfw/command/build-docker command into infra/command and change name to build-docker-image

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoFix CRLF new line (#5959)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 29 Jul 2019 01:28:51 +0000 (10:28 +0900)]
Fix CRLF new line (#5959)

Fix CRLF new line checking error by 19_NN_Compiler_and_Return_SDD.rst

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Rename functions in TemplTensorBuilder (#5943)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 29 Jul 2019 01:21:01 +0000 (10:21 +0900)]
[neurun] Rename functions in TemplTensorBuilder (#5943)

Function `start_lifetime` and `end_lifetime` could be misleading sine
they may or may not do. So this commit renames them to `def_handler` and
`use_handler`, respectively.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoMock ACL NEON nodes for further implementations (#5942)
Nikita Sizov/AI Tools Lab /SRR/Professional/삼성전자 [Mon, 29 Jul 2019 01:07:40 +0000 (04:07 +0300)]
Mock ACL NEON nodes for further implementations (#5942)

* Add mock functions for ACL NEON nodes

Mock: Squeeze, StridedSlice, Transpose, Div, Exp, ReduceMax and Comparison

Signed-off-by: Sizov Nikita <n.sizov@samsung.com>
* Use unused variables in mocked functions

Signed-off-by: Sizov Nikita <n.sizov@samsung.com>
5 years agoRemove .FORMATDENY files from properly formatted directories (#5945)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Mon, 29 Jul 2019 01:07:00 +0000 (04:07 +0300)]
Remove .FORMATDENY files from properly formatted directories (#5945)

Remove.FORMATDENY files from mir_{caffe2,caffe,onnx,tflite}_importer directories which have been recently formatted.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoSupport activation operations for acl neon (#5932)
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 29 Jul 2019 01:00:31 +0000 (10:00 +0900)]
Support activation operations for acl neon (#5932)

This commit supports activation operations for acl neon.
  - Abs, ReLU*, SQRT

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years ago[res/TensorFlowTests] Modify Custom Op user data (#5941)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 29 Jul 2019 00:43:14 +0000 (09:43 +0900)]
[res/TensorFlowTests] Modify Custom Op user data (#5941)

This commit adds `dtype` of custom op, which will be accessed for TypeInference.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years agoDR1 RNN requirements (#5772)
Efimov Alexander/AI Tools Lab/./Samsung Electronics [Mon, 29 Jul 2019 00:26:42 +0000 (09:26 +0900)]
DR1 RNN requirements (#5772)

* DR1 RNN requirements

Add requirements related to RNN networks
remove redundant requirements

Signed-off-by: Efimov Alexander <a.efimov@samsung.com>
5 years agoDr1 development document (#5646)
Efimov Alexander/AI Tools Lab/./Samsung Electronics [Mon, 29 Jul 2019 00:24:50 +0000 (09:24 +0900)]
Dr1 development document (#5646)

* Updated SDD document for DR1

- Added more tasks
- Changed formatiing of task table
- Responsible person is changed

* change development plan, fix responcibilities, remove NNCC repo mentions
* add control flow related item
* remove redundant tizen requirements
* relax control flow requirements
* fix naming of components

Signed-off-by: Efimov Alexander <a.efimov@samsung.com>
5 years ago[moco-tf] remove ModelMetainfo.proto (#5940)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 29 Jul 2019 00:09:56 +0000 (09:09 +0900)]
[moco-tf] remove ModelMetainfo.proto (#5940)

This file, a schema for user input, is now moved into tf2tflite.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[moco-tf] Implement shape inference for TFReshape (#5914)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Sun, 28 Jul 2019 23:52:34 +0000 (08:52 +0900)]
[moco-tf] Implement shape inference for TFReshape (#5914)

This commit implements shape inference for TFReshape on fix_shape().

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[loco] Support traversal to Graph from Node (#5910)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Sun, 28 Jul 2019 23:37:41 +0000 (08:37 +0900)]
[loco] Support traversal to Graph from Node (#5910)

This commit extends Node class to provide "graph" method which returns
a pointer to associated Graph object.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years agoExtract importers out of nnc (#5900)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 26 Jul 2019 11:31:51 +0000 (14:31 +0300)]
Extract importers out of nnc (#5900)

Move sources of Caffe, Caffe2, TensorFlow Lite, ONNX importers from `nnc` to top-level `compiler` directory.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[mir2loco] Support Pooling operations (#5846)
Павел Ильютченко/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 26 Jul 2019 11:25:42 +0000 (14:25 +0300)]
[mir2loco] Support Pooling operations (#5846)

* Support Pooling operation and tests for 2 ops

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
5 years ago[nnc] Cleanup after separating importers (#5902)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 26 Jul 2019 10:55:34 +0000 (13:55 +0300)]
[nnc] Cleanup after separating importers (#5902)

Importers have been extracted into independent components. This is a follow-up cleanup.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoReformat mir importers (#5904)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 26 Jul 2019 10:54:01 +0000 (13:54 +0300)]
Reformat mir importers (#5904)

Reformat code according to coding style.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[neurun] Fix load_model_from_file (#5928)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 26 Jul 2019 09:57:14 +0000 (18:57 +0900)]
[neurun] Fix load_model_from_file (#5928)

load_load_from_file should load model from {nnpackage path + model path}.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
5 years ago[neurun] Remove unnecessary methods of ITensorBuilder (#5923)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 26 Jul 2019 07:32:40 +0000 (16:32 +0900)]
[neurun] Remove unnecessary methods of ITensorBuilder (#5923)

Remove unnecessary methods of ITensorBuilder, cpu/TensorBuilder and
TemplTensorBuilder

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years ago[neurun] Introduce OperationIndexMap as Operand has (#5918)
Дилшоджон Умронхонович Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 26 Jul 2019 06:30:44 +0000 (15:30 +0900)]
[neurun] Introduce OperationIndexMap as Operand has (#5918)

There are a lot of use `std::unordered_map<model::OperationIndex, T>`
So, introducing map alias like Operand has is logical

Signed-off-by: Dilshodzhon Poshshoev <d.poshshoev@samsung.com>
5 years ago[neurun] Generalize ExecutionObserver interface design (#5867)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 26 Jul 2019 05:36:37 +0000 (14:36 +0900)]
[neurun] Generalize ExecutionObserver interface design (#5867)

The current design declares "uploadExecTime" which is a bit specific to
dedicated observer implementation.

This commit revises ExecutionObserver to hook on the whole model execution,
and ProfileObserver to invoke "uploadExecTime" when model execution
ends.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] Apply plier-tf to AvgPoolCanonicalizer (#5921)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 26 Jul 2019 05:30:04 +0000 (14:30 +0900)]
[moco-tf] Apply plier-tf to AvgPoolCanonicalizer (#5921)

This commit applies plier-tf to AvgPoolCanonicalizer.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Applying plier-tf (#5891)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 26 Jul 2019 05:00:11 +0000 (14:00 +0900)]
[moco-tf] Applying plier-tf (#5891)

This commit applies `plier-tf` into `moco-tf`.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years agoRemove exceptional case in python style check (#5917)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 26 Jul 2019 04:19:56 +0000 (13:19 +0900)]
Remove exceptional case in python style check (#5917)

Remove skip compiler directory check on python style check

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Enable handling subtensors for notify{First|Last}Uses (#5915)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 26 Jul 2019 04:18:21 +0000 (13:18 +0900)]
[neurun] Enable handling subtensors for notify{First|Last}Uses (#5915)

Enable handling subtensors for notify{First|Last}Uses by preVisit and
postVisit.

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years agoFormat check for diff file with master (#5916)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 26 Jul 2019 03:46:34 +0000 (12:46 +0900)]
Format check for diff file with master (#5916)

This commit allow user to check format diff files with master via CHECK_DIFF_ONLY environment variable
Use this feature on pre-push githook

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Selective HI_PERF_CPU backend build (#5913)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 26 Jul 2019 03:00:17 +0000 (12:00 +0900)]
[neurun] Selective HI_PERF_CPU backend build (#5913)

This commit allows users to diable HI_PERF_CPU backend build via
BUILD_NEURUN_HI_PERF_CPU_BACKEND option.

Please note that this commit does not change the default behavior.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[neurun] Apply acl internal buffer manager into KernelGenerator (#5908)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 26 Jul 2019 02:45:07 +0000 (11:45 +0900)]
[neurun] Apply acl internal buffer manager into KernelGenerator (#5908)

Apply acl internal buffer manager into KernelGenerator on cl/neon

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years ago[neurun] Append acl_memory_manager method into TemplTensorBuilder (#5887)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 26 Jul 2019 00:08:03 +0000 (09:08 +0900)]
[neurun] Append acl_memory_manager method into TemplTensorBuilder (#5887)

Append acl_memory_manager method into TemplTensorBuilder to be used for
MemoryManager's internal buffer manager

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years ago[neurun] Revise planTensors of Linear (#5882)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 26 Jul 2019 00:07:53 +0000 (09:07 +0900)]
[neurun] Revise planTensors of Linear (#5882)

Revise planTensors of Linear on neurun

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years ago[mocotest-tf] Add TF_SMALL_NET_0005 test (#5888)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 25 Jul 2019 23:38:44 +0000 (08:38 +0900)]
[mocotest-tf] Add TF_SMALL_NET_0005 test (#5888)

This test is consisted of `Reshape` and `Shape` operation.

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
5 years ago[loco] Fix canonical node list (#5893)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 23:29:57 +0000 (08:29 +0900)]
[loco] Fix canonical node list (#5893)

This will fix canonical node list to match alphabetical order of foward

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[tf2tflite] Enable Relu6_000 test (#5886)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 08:59:11 +0000 (17:59 +0900)]
[tf2tflite] Enable Relu6_000 test (#5886)

tf2tflite test framework now checks Relu6_000 automatically.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] Introduce AvgPool Canonicalizer (#5819)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 08:39:50 +0000 (17:39 +0900)]
[moco-tf] Introduce AvgPool Canonicalizer (#5819)

This will introduce AvgPool node Canonicalizier

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[neurun] Append preVisit and postVisit to ITensorBuilder (#5879)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 25 Jul 2019 08:33:03 +0000 (17:33 +0900)]
[neurun] Append preVisit and postVisit to ITensorBuilder (#5879)

Append preVisit and postVisit to ITensorBuilder for #5642

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years ago[exo-tflite] Support ReLU6 (#5884)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 07:23:11 +0000 (16:23 +0900)]
[exo-tflite] Support ReLU6 (#5884)

TFLExporter is now able to export ReLU6 node as T/F Lite RELU6
operation.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] Revise Indentity GraphBuilder (#5855)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 06:41:24 +0000 (15:41 +0900)]
[moco-tf] Revise Indentity GraphBuilder (#5855)

This will revise Identity node GraphBuilder to either convert to TF or Canonical dialect by a knob

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[plier/tf] Move more func from current version of moco-tf Convert.h (#5869)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 06:21:58 +0000 (15:21 +0900)]
[plier/tf] Move more func from current version of moco-tf Convert.h (#5869)

* [plier/tf] Move more func from current version of moco-tf Convert.h

This commit moves more func from current version of moco-tf Convert.h.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* add require(loco). format checked

* error when trying to convert DT_BOOL to loco::DataType

* fix wrong modification

* fix wrong modification

5 years ago[loco] Support ReLU6 type inference (#5877)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 06:21:25 +0000 (15:21 +0900)]
[loco] Support ReLU6 type inference (#5877)

Now, type inference framework supports ReLU6 node.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[plier/tf] remove macro that does not fit with plier-tf (#5871)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 06:04:20 +0000 (15:04 +0900)]
[plier/tf] remove macro that does not fit with plier-tf (#5871)

Remove a macro, which has no relation with TF and plier-tf.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[res] Add test for endmost part of Inception V3 (#5834)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 25 Jul 2019 05:54:39 +0000 (14:54 +0900)]
[res] Add test for endmost part of Inception V3 (#5834)

* [res] Add test for endmost part of Inception V3

This commit adds test case for 'the Epilogue', or the endmost part of
Inception V3 comprised of Squeeze, Reshape, Shape and Softmax.

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

5 years ago[neurun] Make visitor for Subgraph override in KernelGenerator (#5876)
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 25 Jul 2019 05:30:28 +0000 (14:30 +0900)]
[neurun] Make visitor for Subgraph override in KernelGenerator (#5876)

Make visitor for Subgraph override in KernelGenerator for preVisit and
postVisit in the future

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years ago[moco-tf] Rename file and GraphBuilder name for MaxPool (#5851)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 05:30:20 +0000 (14:30 +0900)]
[moco-tf] Rename file and GraphBuilder name for MaxPool (#5851)

This will rename file and GraphBuilder name for MaxPool to follow TensorFlow GraphDef

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years agoCode refactoring Scheduler except makeRank method (#5822)
Дилшоджон Умронхонович Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 05:29:02 +0000 (14:29 +0900)]
Code refactoring Scheduler except makeRank method (#5822)

Split Scheduler::scheduleNode into smaller ones, created
const Graph *_graph, did const some temp variables

Signed-off-by: Dilshodzhon Poshshoev <d.poshshoev@samsung.com>
5 years ago[exo-tflite] Remove unused type inference code (#5859)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 05:28:17 +0000 (14:28 +0900)]
[exo-tflite] Remove unused type inference code (#5859)

exo-tflite now reuses type inference framework that loco provides.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] Introduce TFSqueeze dialect node (#5843)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 25 Jul 2019 05:27:11 +0000 (14:27 +0900)]
[moco-tf] Introduce TFSqueeze dialect node (#5843)

* [moco-tf] Introduce TFSqueeze dialect node

This commit introduces TFSqueeze dialect node.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Use 'squeeze_dims' instead of 'axis'

5 years ago[neurun] Add custom operation node (#5598)
Vladimir Plazun/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 05:25:16 +0000 (08:25 +0300)]
[neurun] Add custom operation node (#5598)

Used as a base node for all custom operations. For general design see at #5411

Signed-off-by: Vladimir Plazun <v.plazun@samsung.com>
5 years ago[neurun] Remove unused prepare method (#5850)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 05:01:30 +0000 (14:01 +0900)]
[neurun] Remove unused prepare method (#5850)

IFunction interface declares prepare method, but no code invokes this
method.

Let's remove this used prepare method.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years agoRemain unified format check script only (#5861)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 04:48:13 +0000 (13:48 +0900)]
Remain unified format check script only (#5861)

Remove nncc/nnfw format check script
Remove all .FORMATCHECKED
Add .FORMATDENY into some directories to ignore format check

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[moco-tf] Run for active nodes for FixPad and FixShape (#5862)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 04:26:56 +0000 (13:26 +0900)]
[moco-tf] Run for active nodes for FixPad and FixShape (#5862)

This will fix to run for active nodes from outputs for FixPadding and FixShape transformations

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years agoDecouple ONNX frontend (#5755)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 04:22:30 +0000 (07:22 +0300)]
Decouple ONNX frontend (#5755)

Remove dependencies of `ONNXImporterImpl` on other `nnc` components.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoDecouple caffe2 frontend (#5754)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 04:19:43 +0000 (07:19 +0300)]
Decouple caffe2 frontend (#5754)

Remove dependencies of `Caffe2Importer` on other `nnc` components.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoDecouple caffe frontend (#5753)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 04:17:52 +0000 (07:17 +0300)]
Decouple caffe frontend (#5753)

Remove dependencies of `CaffeImporter` on other `nnc` components.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoReplace githook and install (#5860)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 04:10:39 +0000 (13:10 +0900)]
Replace githook and install (#5860)

Replace githook to infra/git-hooks
Replace githook install command to infra/command
Update path in script

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoDecouple TFLite frontend (#5752)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 04:05:04 +0000 (07:05 +0300)]
Decouple TFLite frontend (#5752)

Remove dependencies of `TfliteImporter` on other `nnc` components.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoReformat mir component (#5847)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 03:52:23 +0000 (06:52 +0300)]
Reformat mir component (#5847)

Reformat code according to coding style.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[moco-tf] Introduce TFReshape dialect node (#5863)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 25 Jul 2019 03:51:56 +0000 (12:51 +0900)]
[moco-tf] Introduce TFReshape dialect node (#5863)

This commit introduces TFReshape dialect node.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[mir2loco] Introduce MIR to Loco transformer (#5751)
Павел Ильютченко/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 03:50:09 +0000 (06:50 +0300)]
[mir2loco] Introduce MIR to Loco transformer (#5751)

* Implemented core structure
* Supported 2 ops: Input and Output
* Unit test for input and output operations

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
5 years ago[moco-tf] Introduce IdentityCanonicalizer (#5854)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 02:46:46 +0000 (11:46 +0900)]
[moco-tf] Introduce IdentityCanonicalizer (#5854)

* [moco-tf] Introduce IdentityCanonicalizer

This will introduce Identity node Canonicalizer

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

* another fix comment

5 years agoUnify PermuteLayer::runTempl and use Object::access (#5778)
Дилшоджон Умронхонович Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 25 Jul 2019 02:12:18 +0000 (11:12 +0900)]
Unify PermuteLayer::runTempl and use Object::access (#5778)

For all types of PermuteNode::Type, runTempl does the same calculation.
The only difference is for rank==4. Unify it and move switch for type
into rank==4 case

Signed-off-by: Dilshodzhon Poshshoev <d.poshshoev@samsung.com>
5 years agoFix format script bug (#5858)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 02:04:15 +0000 (11:04 +0900)]
Fix format script bug (#5858)

Fix format script bug checking environment variable

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoChange tab size in format script (#5852)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 25 Jul 2019 01:30:21 +0000 (10:30 +0900)]
Change tab size in format script (#5852)

Fix tab size to 2 in format script
Same tab size with other scripts in infra/nncc/command and infra/nnfw/command

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[moco-tf] Fix test of fix shape and pad transform (#5848)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 01:17:36 +0000 (10:17 +0900)]
[moco-tf] Fix test of fix shape and pad transform (#5848)

This will set outputs of test graph for FixShape and FixPadding transform

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Constant Folding Transform for Canonical Dialect (#5759)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 00:59:31 +0000 (09:59 +0900)]
[moco-tf] Constant Folding Transform for Canonical Dialect (#5759)

* [moco-tf] Constant Folding Transform for Canonical Dialect

This commit adds constant folding transform for canonical dialect, plus a test and build files.

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

5 years agoUpdate ExecutorFactory comments (#5844)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 25 Jul 2019 00:30:17 +0000 (09:30 +0900)]
Update ExecutorFactory comments (#5844)

Update outdated comments in `ExecutorFactory.cc`.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[moco-tf] Remove knob condition for Canonicalizer (#5833)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 25 Jul 2019 00:21:09 +0000 (09:21 +0900)]
[moco-tf] Remove knob condition for Canonicalizer (#5833)

This will remove condition with knob for checking for specific TF dialect node
With this change there should be no TF dialects after Canonicalization

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