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>
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>
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>
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>
[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>
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>
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>
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>
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
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>
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>
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>
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>
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>
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>
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>
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>
[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>
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>
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>
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>
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>
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>
[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>
[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>
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>
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>
[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>
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>
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>
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>
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>
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>
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
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>
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>
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>
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>
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>
[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>
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>
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>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 09:39:18 +0000 (18:39 +0900)]
[moco-tf] Remove unused IR headers (#8541)
This will remove not used anymore IR headers that only redirects to moco IR
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 09:18:08 +0000 (18:18 +0900)]
[exo] Revising setCandidate() of FuseBiasAddPass (#8538)
setCandidate() of FuseBiasAddPass was revised to support FormerT and LatterT, and removed names related to conv2d.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 09:03:20 +0000 (18:03 +0900)]
[moco-tf] Remove unused GraphBuilder redirect headers (#8534)
This will remove not used anymore GraphBuilder redirect header files
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 09:02:59 +0000 (18:02 +0900)]
[circle-inspect] Rename "Load.cpp" as "Model.cpp" (#8535)
"Load.cpp" includes the implementation of "Model.h".
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 08:26:53 +0000 (17:26 +0900)]
[moco-tf] Remove Importer.h (#8533)
This will remove not used anymore Importer.h file
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 08:23:13 +0000 (17:23 +0900)]
[exo] moving dynamic cast check in setCandidate into visitors (#8531)
This moves dynamic cast check in setCandidate into visitors.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 08:20:25 +0000 (17:20 +0900)]
[circle-inspect] Dump circle file (#8529)
* [circle-inspect] Dump circle file
This will enable dump circle file operators
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* remove unused headers
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 28 Oct 2019 08:14:00 +0000 (17:14 +0900)]
[tfl-inspect] Rename "Load.cpp" as "Model.cpp" (#8528)
"Load.cpp" includes the implementation of "Model.h".
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 08:12:40 +0000 (17:12 +0900)]
[moco-tf] Remove Dialect files (#8527)
This will remove not used anymore Dialect folder files
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 28 Oct 2019 08:10:20 +0000 (17:10 +0900)]
Introduce tf2tflite integration test (#8523)
This commit introduces tf2tflite integration test:
tf2tflite-integration-test directory, and tf2tflite_integration_test
test executable
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 08:10:04 +0000 (17:10 +0900)]
[exo] revising Fuse class to use FormerT and LatterT (#8530)
This revises Fuse class in FuseBiasAddPass.cpp to use FormerT and LatterT.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
채성우/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 28 Oct 2019 07:39:19 +0000 (16:39 +0900)]
[infra/cmake] add TensorFlowSource 1.13.1 (#8519)
This commit adds TensorFlowSource 1.13.1
Signed-off-by: seongwoo <sw4670.chae@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 07:38:40 +0000 (16:38 +0900)]
[circle-inspect] Introduce Load Reader and Model (#8525)
* [circle-inspect] Introduce Load Reader and Model
This will introduce circle model Load and Reader using mio-circle
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix comment
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 28 Oct 2019 07:35:10 +0000 (16:35 +0900)]
[tf2circle] Set shape information in signature at tf2circle (#8484)
Until now, signature in `tf2circle` didn't have shape information.
This commit will set the shape information in signature at `tf2circle`
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 07:13:40 +0000 (16:13 +0900)]
[moco-tf] Update to use moco TFDialect (#8522)
This will update to use moco TFDialect in src folder
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 06:50:30 +0000 (15:50 +0900)]
[exo] rewriting FuseBiasAddPass to use Mixin<Bias> (#8507)
* [exo] FuseBiasAddPass
This code is modified to use Mixin<Bias> instead of TFLConv2D. (This commit is a part of modification.)
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* Former -> FormerT
* Fix get_former()
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 06:36:07 +0000 (15:36 +0900)]
[moco-tf] Update to use moco TFDialect (#8514)
This will update to use TFDialect from moco project
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 06:35:47 +0000 (15:35 +0900)]
[circle_inspect] Initial circle-inspect project (#8513)
This will introduce circle-inspect project to inspect Circle model file
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 06:35:30 +0000 (15:35 +0900)]
[moco-tf] Update Transform header with moco (#8512)
This will update Transforms header to include individual Passes from moco and delete not used anymore redirect header files
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 06:23:18 +0000 (15:23 +0900)]
[tf2nnpk] check existence and provide help (#8499)
It will show help with `--help`.
It will check existence of flatc, graphdef and info file.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 28 Oct 2019 04:59:22 +0000 (13:59 +0900)]
Make to support Pack op for acl neon (#8506)
This commit makes to support Pack op for acl neon.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 28 Oct 2019 04:52:02 +0000 (13:52 +0900)]
[ncnn-backend] Enable NCNN backend test (#8482)
Test ncnn backend - transpose conv only
(run with cpu backend test job)
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 04:09:47 +0000 (13:09 +0900)]
[moco-tf] Use moco Importer (#8504)
This will update to use moco Importer header
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
[neurun] Change Gather axis param to int (#8460)
* Changed Gather axis param from OperandIndex to int
Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 03:57:48 +0000 (12:57 +0900)]
[moco_tf] Remove ResolveFusedBatchNorm test (#8502)
This will remove ResolveFusedBatchNorm test as the Pass itself has migrated to moco-pass
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 03:46:54 +0000 (12:46 +0900)]
[moco-tf] Use mono namespace of Passes (#8500)
This will update to use moco namespace of Passes
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 03:19:12 +0000 (12:19 +0900)]
[tfl-inspect] Introduce dump operators (#8493)
* [tfl-inspect] Introduce dump operators
This will introduce DumpOperators and use this for dumping operators in the model file
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* apply comments
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 28 Oct 2019 01:53:51 +0000 (10:53 +0900)]
[tf2nnpkg] Exit immediately on fail (#8468)
Exit immediately if one of called script or binary (tf2tflite, tflite2circle, model2nnpkg) failed
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 01:51:30 +0000 (10:51 +0900)]
[moco-tf] Redirect to use moco Pass (#8492)
This will redirect to use moco Pass and delete migrated Pass implementions
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 28 Oct 2019 01:11:14 +0000 (10:11 +0900)]
Make to support Pack op for acl_cl (#8490)
This commit makes to support Pack op for acl_cl.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 28 Oct 2019 01:09:35 +0000 (10:09 +0900)]
Change calculating offset of Readers and Views (#8409)
This commit changes calculating offset of Readers and Views to use strides.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 28 Oct 2019 00:55:12 +0000 (09:55 +0900)]
[nnpkg_test] nnpkg_test list accepts wildcard (#8486)
- You can use * in list.
- nnpkg_test list is updated.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 27 Oct 2019 21:48:20 +0000 (06:48 +0900)]
[exo] revising comment for the target nodes of FusedBiasAddPass (#8477)
Comment for the target nodes of FusedBiasAddPass was revised.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 27 Oct 2019 21:45:42 +0000 (06:45 +0900)]
[exo] mixin for nodes that have bias input and fused activation function (#8476)
* [exo] mixin for nodes that have bias input and fused activation function
This adds a mixin for nodes that have bias input and fused activation function.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* remove inheritance structure of mixin
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 27 Oct 2019 21:28:16 +0000 (06:28 +0900)]
[exo] renaming template type and node name (#8483)
Instead of trying to imply real type of template type,
general T is used and static_assert was added to explicitly show allowed type.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 27 Oct 2019 21:26:44 +0000 (06:26 +0900)]
[tfl-inspect] Introduce Reader (#8461)
* [tfl-inspect] Introduce Reader
This will introduce tflite file Reader class and required methods
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use tflinspect namespace
* move Model inside src
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 27 Oct 2019 21:25:58 +0000 (06:25 +0900)]
[moco] Introduce Pass (#8466)
* [moco] Introduce Pass
This will introduce Pass that provides TF dialect pass, transformation and optimization
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix typo
[nnc] Extract implementations of activation functions (#8487)
Extract implementation of activation function to cpp files.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 25 Oct 2019 10:29:56 +0000 (19:29 +0900)]
Introduce Pack op (#8355)
* Introduce Pack op
This commit introduces Pack op into frontend.
- Introduce PackNode into neurun
- Introduce Pack op into frontend
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
* Add NYI runtime_error for pack op of base_loader
This commit adds NYI runtime_error for pack op of base_loader.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 25 Oct 2019 08:12:49 +0000 (17:12 +0900)]
Update nnpkg_test list (#8481)
This commit updates nnpkg_test's test coverage.
Note that unit test for 'Concat' and 'DepthwiseConv2D' have been changed
their names:
- UNIT_Concat_* -> UNIT_ConcatV2_*
- UNIT_DepthwiseConv2D_* -> UNIT_DepthwiseConv2dNative_*
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 25 Oct 2019 08:00:25 +0000 (17:00 +0900)]
[android] Introduce android_benchmark_app (#8380)
* [android] Introduce android_benchmark_app
This commit introduces an Android benchmark app for easy testing with an
actual Android visual app rather than running on shell.
The app contains minimal UI which consist of 2 buttons and few text
views to show results.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
* Fixes
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 25 Oct 2019 07:53:28 +0000 (16:53 +0900)]
[CI] Infra scripts for nncc package and nnpkg test (#8471)
CI infra scripts
- nncc build and resouce collection scripts
- Update build_nnpkg.sh to install nncc package
- Update nncc-tc-to-nnpk-tc.sh to handle pb file and use tf2nnpkg
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 25 Oct 2019 07:51:25 +0000 (16:51 +0900)]
[exo] renaming a method to a bit more detaild name (#8478)
This renames `create_fused_bias()` to `create_fused_bias_const()`.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 25 Oct 2019 07:40:06 +0000 (16:40 +0900)]
[exo] refactoring: rearranging if.. else.. in FuseBiasAddPass (#8479)
This re-arranges `if (abnormal case) else if (normal case)` to `if (normal case) else if (abnormal case)` for better readability.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 25 Oct 2019 06:24:23 +0000 (15:24 +0900)]
'create-package' with relative install prefix (#8443)
* 'create-package' with relative install prefix
This commit allows './nnas create-package' with relative install prefix
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Fix to be relative to user's directory
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 25 Oct 2019 06:22:38 +0000 (15:22 +0900)]
[exo] setting FusedActFunc of TFLAdd and TFLSub with init() function (#8454)
This adds init() method that sets FusedActFunc of TFLAdd and TFLSub
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 25 Oct 2019 03:45:30 +0000 (12:45 +0900)]
[tf2nnpkg] Model name (#8467)
Use model name for circle file and package directory output
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 25 Oct 2019 03:33:14 +0000 (12:33 +0900)]
[nnas] Introduce verify-package command (#8435)
This command allows users to verify NNAS Host Toolchain package. The
current implementation supports very basic checks.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 25 Oct 2019 02:08:27 +0000 (11:08 +0900)]
[nnfw/cmd] Remove alias setting (#8359)
* [nnfw/cmd] Remove alias setting
Remove alias setting in command
Directory alias will be set on Makefile
Add and fix build path checking message
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Fix bug
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 25 Oct 2019 01:30:02 +0000 (10:30 +0900)]
[moco] Remove service from lang (#8465)
This will remove service folder from Lang that exist as separate target
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 25 Oct 2019 01:08:45 +0000 (10:08 +0900)]
[nnpackage_run] Fix time unit shown (#8398)
Fix printed time unit `sec` to `ms`.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 25 Oct 2019 00:55:22 +0000 (09:55 +0900)]
[exo] renaming test name (FuseConv2DBiasAddPassTest to FuseBiasAddPassTest) (#8455)
Test name was renamed from FuseConv2DBiasAddPassTest to FuseBiasAddPassTest.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 25 Oct 2019 00:45:51 +0000 (09:45 +0900)]
[moco] Let Support use Lang (#8462)
This will update Support to use Lang target
- delete migrated TFDataLayout and TFPadding
- use TFDataLayout and TFPadding from Lang
- update cmake build order
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 24 Oct 2019 23:16:17 +0000 (08:16 +0900)]
[res/TensorFlowTests] test models for fusing DepthwiseConv with other ops (#8457)
This adds test models for fusing DepthwiseConv with other ops.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
[mir-tflite-importer] Fix setting quantization only on uint8 tensors (#8458)
* Added data type check for all creations of quantization
Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
[mir-tflite-importer] Fix checking quantized input type (#8456)
* Fixed setting quantization for InputOp then data type isn't UINT8
Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 24 Oct 2019 11:32:04 +0000 (20:32 +0900)]
[nnfw-api] Remove header install by copy (#8376)
Divide nnfw api directory to src and include, and install public headers
Rename nnapi directory and install public header
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 24 Oct 2019 10:21:36 +0000 (19:21 +0900)]
[exo] Convert DepthwiseConv2D applied (#8449)
This commit turns on conversion for DepthwiseConv2D
Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>