박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 1 Aug 2019 05:09:20 +0000 (14:09 +0900)]
[loco] Introduce NodeShape class (#6078)
* [loco] Introduce NodeShape class
This commit introduces NodeShape class which serves as a variant over
various Shape classes.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Fix bugs and add tests
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 05:01:12 +0000 (14:01 +0900)]
[locomotiv] execution for elementwise add (#6079)
* [locomotiv] execution for elementwise add
This will introduce execution for EltwiseAdd node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* validate shape
* remove using namespace
* check if same domain not tensor
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 04:21:40 +0000 (13:21 +0900)]
[moco-tf] Introduce MulCanonicalizer (#6082)
This will introduce Canonicalizer for Mul node. Supports only for Elementwise mul for now.
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 1 Aug 2019 04:08:46 +0000 (13:08 +0900)]
[moco-tf] Simplify TFFormattedGraph using pepper::str (#6080)
This commit simplifies TFFormattedGraph implementation using
"pepper::str" helper.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 04:07:48 +0000 (13:07 +0900)]
[moco-tf] use plier method (#6090)
This will fix to use plier method from old internal method
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 03:34:50 +0000 (12:34 +0900)]
[moco-tf] Import as TFMaxPool with a knob (#6043)
* [moco-tf] Import as TFMaxPool with a knob
This will revise import to load MaxPool node as TFMaxPool with a knob
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use plier and rename variable
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 03:34:33 +0000 (12:34 +0900)]
[tf2tflite] Schema for additional info of Custom Op from user (#5939)
* [tf2tflite] Schema for additional info of Custom Op from user
This is a schema for additional info of Custom Op from user.
It was committed into _moco-tf_ but moved to here.
The name was renamed to CompileInfo.pbtxt (Previous name was Model.MetaInfo.pbtxt), which seems more friendlier. :-)
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* renaming CompileInfo to CustomInfo && move ./proto/CMakeFiles.txt to ./CMakeFiles.txt
* modify DT_INT32 to 15
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 1 Aug 2019 03:34:16 +0000 (03:34 +0000)]
[exo/tflite] prevent ceil_div function from overflow (#6088)
This commit prevents ceil_div function from overflow.
Signed-off-by: seongwoo <sw4670.chae@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 03:25:48 +0000 (12:25 +0900)]
[moco-tf] Logging with TFNodeSummaryBuilderFactory (#6064)
TFNodeSummaryBuilderFactory is used to print all nodes (including unlinked TFNodes)
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 1 Aug 2019 02:30:32 +0000 (11:30 +0900)]
Introduce PermuteFactor into LowerInfo of operation (#6028)
This commit introduces PermuteFactor into LowerInfo of operation.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 1 Aug 2019 02:29:58 +0000 (11:29 +0900)]
Refactor Backend to support sessions (#5894)
Backend had been only one object there cannot be more than one session.
This commit introduces `BackendContext` which depends on a specific
model and `Backend` object represents the backend itself(a singleton).
Resolve #5723
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 1 Aug 2019 01:50:47 +0000 (01:50 +0000)]
[moco/tf] Introduce DepthwiseConv2D Canonicalizer (#6047)
* [moco/tf] Introduce DepthwiseConv2D Canonicalizer
This commit introcude DepthwiseConv2D Canonicalizer to moco-tf.
Signed-off-by: seongwoo <sw4670.chae@samsung.com>
* apply comment.
* add codes for log.
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 1 Aug 2019 01:45:06 +0000 (10:45 +0900)]
[moco-tf] Fix build break (#6087)
This commit fixes build break related with API deisgn changes.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 01:37:38 +0000 (10:37 +0900)]
[moco-tf] Add a complex test case into ConstantFoldingTransform (#5936)
* [moco-tf] Add a complex test case into ConstantFoldingTransform
This commit adds a complex test case into ConstantFoldingTransform
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* using TFNodeSummerBuilderFactory
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 1 Aug 2019 01:36:01 +0000 (10:36 +0900)]
[moco-tf] Transform to resolve wildcard dimesion of TFReshape (#6019)
* [moco-tf] Transform to resolve wildcard dimesion of TFReshape
This commit introduces ResolveReshapeWildcardDim transform, which tries
to resolve wildcard dimension of TFReshape.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Clarify by docs and function name
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 1 Aug 2019 01:35:26 +0000 (10:35 +0900)]
[res] Add small network test case for ResolveReshapeWildcardDim (#6020)
* [res] Add small network test case for ResolveReshapeWildcardDim
This commit adds small network test case for ResolveReshapeWildcardDim.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Edit brief description
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 01:11:29 +0000 (10:11 +0900)]
[moco-tf] Remove functions in Convert.h that were moved to plier-tf (#5935)
* [moco-tf] Remove functions in Convert.h that were moved to plier-tf
This commit removes functions in Convert.h that were moved to plier-tf.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* remaining asDataLayout
* remove unused header
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 1 Aug 2019 00:04:45 +0000 (09:04 +0900)]
[moco-tf] Adding more methods into ModelSignature (#5897)
* [moco-tf] Adding more methods into ModelSignature
This commit adds more methods into ModelSignature to make info from users (regarding custom op) be stored.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* format checked
* use angkor::TensorShape
* renaming customop(..) -> add_customop(..) && Fix comment
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 31 Jul 2019 23:48:16 +0000 (08:48 +0900)]
[loco] Add BiasShape class (#6060)
This commit adds BiasShape class which records the length of bias
values.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 31 Jul 2019 23:33:01 +0000 (08:33 +0900)]
[loco] Introduce EltwiseMul IR (#6055)
This will introduce EltwiseMul IR for Elementwise Multiply operation
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 31 Jul 2019 23:32:38 +0000 (08:32 +0900)]
[moco-tf] Support TFMaxPool for Logging (#6054)
This will add TFMaxPool for TFNodeSummaryBuilder this dump of graph shows TFMaxPool attributes
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 31 Jul 2019 23:29:31 +0000 (08:29 +0900)]
[moco-tf] Update readme (#5870)
* [moco-tf] Update readme
This will update README.md file of moco-tf project
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* some updates
* use requires.cmake
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 31 Jul 2019 22:32:57 +0000 (07:32 +0900)]
[moco-tf] Canonicalize MaxPool (#6046)
This will enable Canonicalization of MaxPool node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
Павел Ильютченко/AI Tools Lab /SRR/Engineer/삼성전자 [Wed, 31 Jul 2019 13:45:33 +0000 (16:45 +0300)]
Fix not consistent input order in concat op (#6067)
* Fixed asserts using check what it is Pull nodes
Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
Nikita Sizov/AI Tools Lab /SRR/Professional/삼성전자 [Wed, 31 Jul 2019 10:07:29 +0000 (13:07 +0300)]
ACL NEON Div Node (#5948)
Add support of DivNode for ACL NEON
Signed-off-by: Sizov Nikita <n.sizov@samsung.com>
Павел Ильютченко/AI Tools Lab /SRR/Engineer/삼성전자 [Wed, 31 Jul 2019 08:58:17 +0000 (11:58 +0300)]
Support Concat operation transformation from mir to loco (#5903)
* Transform ConcatOp to TensorConcat
Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
Павел Ильютченко/AI Tools Lab /SRR/Engineer/삼성전자 [Wed, 31 Jul 2019 08:56:42 +0000 (11:56 +0300)]
[mir2loco] Support Relu operation (#5793)
* Support Relu operation transformation
* Append test for Relu
Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 31 Jul 2019 07:51:07 +0000 (16:51 +0900)]
[plier-tf] Moving moco test cases regarding data_layout (#6061)
Test cases of moco were moved into plier.
These test cases are for two methods that were moved lately from moco to plier.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 31 Jul 2019 07:43:29 +0000 (16:43 +0900)]
[pepper-str] Initial commit (#6056)
This commit adds pepper-str module. "pepper-str" provides "str" helper
which allows users to simulate string interpolation in C++.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 31 Jul 2019 07:34:13 +0000 (16:34 +0900)]
[moco-tf] TFReshape Canonicalization (#6005)
* [moco-tf] TFReshape Canonicalization
This commit implements Canonicalization for TFReshape. Currently it only
supports fixed reshape case.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* TODO comment to support loco ConstGen node
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 31 Jul 2019 06:56:29 +0000 (15:56 +0900)]
Fix invalid params of CLTransposeConv (#6057)
This commit fixes invalid params of CLTansposeConv.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 31 Jul 2019 06:33:31 +0000 (15:33 +0900)]
[loco] Use data type attribute in Graph-level input (#5999)
This commit revises "Pull" node implementation to use Graph-level
"dtype" attribute.
This commit includes some workaround not to impact the current clients.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 31 Jul 2019 06:19:51 +0000 (15:19 +0900)]
[neurun] Rename "make_subgraph" helper (#6042)
Graph::lower internally declares "make_subgraph", but its role is more
than subgraph construction.
This commit renames "make_subgraph" as "append_fresh_single_op_subgraph"
to make it easy to understand its behavior.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 31 Jul 2019 01:20:40 +0000 (10:20 +0900)]
[moco-tf] Support TFAvgPool for Logging (#6045)
This will add TFAvgPool for TFNodeSummaryBuilder thus dump of graph shows TFAvgPool attributes
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 31 Jul 2019 00:41:36 +0000 (09:41 +0900)]
[moco-tf] Canonicalizer for MaxPool (#6027)
This will introduce MaxPoolCanonicalizer to canonicalize MaxPool node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 31 Jul 2019 00:30:12 +0000 (09:30 +0900)]
[loco] Introduce EltwiseAdd IR (#6026)
* [loco] Introduce EltwiseAdd IR
This will introduce EltwiseAdd IR for Elementwise Add operation
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use lhs, rhs
* remove Tensor in comment
DR1 Detailed level documentation (#5896)
* DR1 Detailed level documentation
Add detailed level documentation for DR1 milestone
* update 2019 goals
* initial structure description
* add compiler flow
* add description of components including compiler and runtime
* add description to high level directories, and add line breaks
* remove redundant data, fix references
* fix typos and wording
* DR1 DLD SE member in-charge review and fixes
- Expanded OSS pre-review table with new columns
- Added SW Code Structure using `tree` tool
- Updated Revision history and Table of contents
Signed-off-by: Efimov Alexander <a.efimov@samsung.com>
DR1 sw requirement document fixes (#5992)
* DR1 sw requirement document fixes
- Fix sw requirement document number version
- Fixed non-function requirements numeration
Signed-off-by: Aleksei Kondrashov <a.kondrashov@partner.samsung.com>
Add High Level Documentation (#5665)
* Add High Level Documentation
- Added image of overall architecture of project
- Added image with data flow through the pipeline
- Added Core Structure component
- Added and mapped non-functional requirements
- Updated revision history
* add overview of components
* add initial version of API secton
* add IR description
* add runtime behavior
* add structural description of IRs
* add compiler structure and behavioral description
* add structure diagram about nn runtime
* add constraints and scope items
* add Traceability Matrix
* add table of contents and items required by template
* add description about gray and white modules on component diagram
* update images to eliminate ambiguous interpretation
* change heading, rename nn Compute component
* break sections in more items
* bring back nnapi part
* rename NNFW IR into neurun IR and fix naming of other components
* fix formatting of subsystem description & abbreviations
* remove redundant item from compiler description
* fix typos and wording
Signed-off-by: Efimov Alexander <a.efimov@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 30 Jul 2019 23:39:48 +0000 (08:39 +0900)]
[moco-tf] Fix shape for TFMaxPool (#6025)
This will implement FixShape for TFMaxPool node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 30 Jul 2019 22:17:27 +0000 (07:17 +0900)]
[moco-tf] Fix pad for TFMaxPool (#6023)
This will implement FixPad for TFMaxPool node
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
[mir] Add support for 0-dimensional tensors (#6039)
Fixes compilation error of SuperResolution and StyleTransfer networks.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir/TFLite importer] Cleanup after changing interface (#6036)
* Remove `cleanup` method.
* Move `import` and `createIR` methods into private section.
* Use `make_unique` to deal with `unique_ptr`s.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir/ONNX importer] Cleanup after changing interface (#6035)
* Move `import` and `createIR` methods into private section.
* Use `make_unique` to deal with `unique_ptr`s.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir/Caffe importer] Cleanup after changing interface (#6034)
* Remove `cleanup` method.
* Move `import` and `createIR` methods into private section.
* Use `make_unique` to deal with `unique_ptr`s.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir/Caffe2 importer] Cleanup after changing interface (#6033)
* Remove `cleanup` method.
* Move `import` and `createIR` methods into private section.
* Use `make_unique` to deal with `unique_ptr`s.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 30 Jul 2019 11:49:11 +0000 (20:49 +0900)]
Fix format script diff file only (#6007)
To get correct updated files when master is update after branch out, count new commits after branch out from master and compare with branch out commit
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 30 Jul 2019 11:48:41 +0000 (20:48 +0900)]
[neurun] Introduce "simple" Subgraph emplace method (#6000)
This commit introduces "simple" Subgraph empalce method which does not
allocate a Subgraph internally but requires clients to allocate it
instead.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
[neurun] Code refactoring of Scheduler::makeRank (#5881)
Split Scheduler::makeRank into smaller ones
Signed-off-by: Dilshodzhon Poshshoev <d.poshshoev@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 30 Jul 2019 10:07:41 +0000 (19:07 +0900)]
[moco-tf] ResolveFusedBatchNorm unit test to generate proper log (#6024)
Previously unit test for ResolveFusedBatchNorm printed unknown node
fallback logging. This commit fixes this bug to print proper format
defined by moco-tf.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 30 Jul 2019 10:06:22 +0000 (10:06 +0000)]
[moco/tf] Import TFDepthwiseConv2d Op (#6001)
* [moco/tf] Import TFDepthwiseConv2d Op
This commit is for importing TFDepthwiseConv2d Op.
Signed-off-by: seongwoo sw4670.chae@samsung.com
* This commit makes tensor_content of pbtxt human-readable.
[nnc] Use `importModel` methods instead of `import`, `createIR` and `cleanup` (#5990)
Replace uses of `import`, `createIR` and `cleanup` with newly introduced `importModel`.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
DR1 test document (#5890)
* DR1 test document
- Introduced STD for DR1 milestone
- According to new requirements new tests are added
Signed-off-by: Efimov Alexander <a.efimov@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 30 Jul 2019 08:52:30 +0000 (17:52 +0900)]
Fix bug in ExecInstance (#6017)
Delete compiler generated by new
Fix typo input buffer name in twoThreads test
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 30 Jul 2019 08:51:45 +0000 (17:51 +0900)]
Replace pylint command (#6002)
Replace infra/nnfw/command/pylint command into infra/command
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Nikita Sizov/AI Tools Lab /SRR/Professional/삼성전자 [Tue, 30 Jul 2019 08:03:28 +0000 (11:03 +0300)]
ACL NEON Comparison Node (#5951)
Add support of ComparisonNode for ACL NEON
Signed-off-by: Sizov Nikita <n.sizov@samsung.com>
Nikita Sizov/AI Tools Lab /SRR/Professional/삼성전자 [Tue, 30 Jul 2019 08:01:25 +0000 (11:01 +0300)]
ACL NEON SqueezeNode (#5944)
Add support of SqueezeNode for ACL NEON
Signed-off-by: Sizov Nikita <n.sizov@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 30 Jul 2019 07:55:32 +0000 (16:55 +0900)]
[moco-tf] Reshape shape inference rejects dynamic case (#6013)
Previously shape inference stage for TFReshape node was erroneously
done for wildcard dimension as well. This commit fixes this bug to
reject dynamic reshape case.
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 30 Jul 2019 07:52:30 +0000 (16:52 +0900)]
Transpose convolution valid padding test (#5998)
* Transpose convolution valid padding test
- Add transpose convolution valid padding test
- Fix bug transpoe valid padding
- Update comment
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Update comment
Fix goals for 2019 (#5994)
Fix goals for 2019 in SRS and SDD documents
Signed-off-by: Efimov Alexander <a.efimov@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 30 Jul 2019 07:21:41 +0000 (16:21 +0900)]
Add gtest to prepare multithread execution instance (#5898)
Introduce gtest to prepare multithread execution instance using one executor
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
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>
[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>
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>
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>
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>
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>
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>
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>
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
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>
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>
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>
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>
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>
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>
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>
[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>
[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>
[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>
[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>
[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>
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>
[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>
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>
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>
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>
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>
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
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
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>
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>
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>
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.
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>
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>
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>
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>
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>
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>