platform/core/ml/nnfw.git
5 years ago[locop] Introduce FormattedTensorShape (#6399)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 9 Aug 2019 01:17:47 +0000 (10:17 +0900)]
[locop] Introduce FormattedTensorShape (#6399)

This commit introduces FormattedTensorShape class which dumps the
content of a given TensorShape into std::ostream.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years agoRevise cpu backend cmake (#6401)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 9 Aug 2019 01:16:05 +0000 (10:16 +0900)]
Revise cpu backend cmake (#6401)

- cpu backend, hi-perf-cpu backend
- Set strict build option: nnfw_common

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[moco-tf] Turn off ImportAsTFConcatV2 (#6421)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 9 Aug 2019 01:02:38 +0000 (10:02 +0900)]
[moco-tf] Turn off ImportAsTFConcatV2 (#6421)

This will turn off ImportAsTFConcatV2 knob before applying VariadicArityNode to TFConcatV2

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Import Shape operation (#6402)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 9 Aug 2019 00:56:46 +0000 (09:56 +0900)]
[moco-tf] Import Shape operation (#6402)

This commit implements import stage for Shape operation

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[nnc] Support for binary elementwise operations (#6408)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 8 Aug 2019 15:30:10 +0000 (18:30 +0300)]
[nnc] Support for binary elementwise operations (#6408)

* Add implementations of binary elementwise operations to backends.
* Switch backend unittests to new ops.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[mir_onnx] Improve core structure of ONNX importer (#6297)
Павел Ильютченко/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 8 Aug 2019 14:03:16 +0000 (17:03 +0300)]
[mir_onnx] Improve core structure of ONNX importer (#6297)

* Fix right initializer loading
* Fix output creation
* Other fixes

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
5 years ago[mir] Move operations.lst.h into include/mir (#6405)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 8 Aug 2019 12:07:11 +0000 (15:07 +0300)]
[mir] Move operations.lst.h into include/mir (#6405)

Move `include/mir/ops/operations.lst.h` -> `include/mir/Operations.inc` so that it can be found more easily.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoUpdate skiplist for acl neon backend (#6396)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 09:53:04 +0000 (18:53 +0900)]
Update skiplist for acl neon backend (#6396)

Update nnapi generated test for acl neon backend
- Enable validation test
- Enable trivial test except broadcast mul
- Enable depth-conv
- Enable comparison
- Enable remain fc tests
- Enable remain reshape tests

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoSupport async execution in NNAPI (#6395)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 09:52:39 +0000 (18:52 +0900)]
Support async execution in NNAPI (#6395)

Implement Event class to support async execution
Use internal async execution function for startCompute and wait

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoRevise acl cl backend cmake (#6391)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 09:51:29 +0000 (18:51 +0900)]
Revise acl cl backend cmake (#6391)

- Set strict build option: nnfw_common
- Remove needless link
- Update header include for acl common

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoRename gather implementation in arm-compute-ex (#6388)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 08:55:46 +0000 (17:55 +0900)]
Rename gather implementation in arm-compute-ex (#6388)

Rename CLGather and CLGatherKernel to CLGatherEx and CLGatherExKernel because acl 19.05 already have same name layer & kernel

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoRevise acl neon backend cmake (#6393)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 08:46:19 +0000 (17:46 +0900)]
Revise acl neon backend cmake (#6393)

- Set strict build option: nnfw_common
- Remove needless link
- Update header include for acl common

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[moco-tf] relocate input method in Concat import (#6394)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 07:58:52 +0000 (16:58 +0900)]
[moco-tf] relocate input method in Concat import (#6394)

This will relocate input method in Concat after class declaration to put anonymous codes together

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Introduce ResolveConstantShape transform (#6379)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 07:43:00 +0000 (16:43 +0900)]
[moco-tf] Introduce ResolveConstantShape transform (#6379)

* [moco-tf] Introduce ResolveConstantShape transform

This commit introduces ResolveConstantShape transform, which is
responsible to replace determined TFShape node with TFConst.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Separate out condition check stage

5 years ago[loco] Update index on link (#6381)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 07:42:30 +0000 (16:42 +0900)]
[loco] Update index on link (#6381)

Both "link" helpers now update the index of a corresponding node.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] TFShape shape inference (#6373)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 07:42:02 +0000 (16:42 +0900)]
[moco-tf] TFShape shape inference (#6373)

This commit implements TFShape node's shape inference.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[exo-tflite] Use shared shape inference framework by default (#6390)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 07:41:42 +0000 (16:41 +0900)]
[exo-tflite] Use shared shape inference framework by default (#6390)

This commit makes exo-tflite to use the shape inference framework
implemented in loco core by default.

Please note that it is possible to turn off this feature via setting
EXOTFLITE_USE_LOCO_SHAPE_INFERENCE as 0.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[loco] Simplify ReLU Graph Testcase (#6392)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 07:39:37 +0000 (16:39 +0900)]
[loco] Simplify ReLU Graph Testcase (#6392)

Let's simplify ReLU Graph Testcase using GraphBuilder.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] Revise sanity check for squeeze dim (#6370)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 07:39:29 +0000 (16:39 +0900)]
[moco-tf] Revise sanity check for squeeze dim (#6370)

This commit would throw error for invalid squeeze dimension.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[res] Small network mocking the Epilogue (#6375)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 07:23:31 +0000 (16:23 +0900)]
[res] Small network mocking the Epilogue (#6375)

This commit adds TF_SMALL_NET_0014, a mock Epilogue network using ReLU
instead of Softmax

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[neurun] Nit : fix OperandNode::BG_COLORS (#6384)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 07:04:05 +0000 (16:04 +0900)]
[neurun] Nit : fix OperandNode::BG_COLORS (#6384)

`BG_COLOR` is a static member of `Node` but it was refereced with a
subclass without a reason.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[moco-tf] relocate MaxPool local class into anonymous (#6385)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 06:56:18 +0000 (15:56 +0900)]
[moco-tf] relocate MaxPool local class into anonymous (#6385)

This will relocate local class of MaxPool import into anonymous namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] relocate Mul local class into anonymous (#6386)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 06:55:53 +0000 (15:55 +0900)]
[moco-tf] relocate Mul local class into anonymous (#6386)

This will relocate local class of Mul import into anonymous namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] relocate Relu local class into anonymous (#6387)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 06:55:34 +0000 (15:55 +0900)]
[moco-tf] relocate Relu local class into anonymous (#6387)

This will relocate local class of Relu import into anonymous namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] relocate Relu6 local class into anonymous (#6389)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 06:52:39 +0000 (15:52 +0900)]
[moco-tf] relocate Relu6 local class into anonymous (#6389)

This will relocate local class of Relu6 import into anonymous namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[loco] Simplify Identity Graph Testcase with Graph Builder (#6344)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 05:42:38 +0000 (14:42 +0900)]
[loco] Simplify Identity Graph Testcase with Graph Builder (#6344)

* [loco] Simplify Identity Graph Testcase with Graph Builder

This commit simplifies Identity Graph Testcase with GraphBuilder.

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

5 years ago[moco-tf] relocate Identity local class into anonymous (#6382)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 05:40:14 +0000 (14:40 +0900)]
[moco-tf] relocate Identity local class into anonymous (#6382)

This will relocate local class of Identity import into anonymous namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Switch on for ImportAsTFConcatV2 (#6380)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 05:24:05 +0000 (14:24 +0900)]
[moco-tf] Switch on for ImportAsTFConcatV2 (#6380)

This will switch on for ImportAsTFConcatV2 in the knob

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Canonicalize ConcatV2 node (#6377)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 04:30:28 +0000 (13:30 +0900)]
[moco-tf] Canonicalize ConcatV2 node (#6377)

This will enable Canonicalization of TFConcatV2 to canonical TensorConcat

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years agoFix acl backend test fail (#6376)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 04:19:50 +0000 (13:19 +0900)]
Fix acl backend test fail (#6376)

Fix acl cl backend test fail

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[loco] Add index getter (#6367)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 04:19:19 +0000 (13:19 +0900)]
[loco] Add index getter (#6367)

This commit extends GraphInput and GraphOutput class with "index"
getter.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[mocotest-tf] Enable Squeeze_002 (#6369)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 04:12:52 +0000 (13:12 +0900)]
[mocotest-tf] Enable Squeeze_002 (#6369)

This commit would enable squeeze network test with negative squeeze
dimension.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] relocate FusedBatchNorm local class into anonymous (#6374)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 04:03:59 +0000 (13:03 +0900)]
[moco-tf] relocate FusedBatchNorm local class into anonymous (#6374)

This will relocate local class of FusedBatchNorm import into anonymous namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] TFShape summary (#6372)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 04:00:26 +0000 (13:00 +0900)]
[moco-tf] TFShape summary (#6372)

This commit adds TFShape node summary.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Introduce ConcatV2Canonicalizer (#6360)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 03:54:17 +0000 (12:54 +0900)]
[moco-tf] Introduce ConcatV2Canonicalizer (#6360)

* [moco-tf] Introduce ConcatV2Canonicalizer

This will introduce ConcatV2Canonicalizer that converts TFConcatV2 to canonical TensorConcat node

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

* remove unused header

5 years agoFix several warnings (#6242)
Vladimir Plazun/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 8 Aug 2019 03:42:02 +0000 (06:42 +0300)]
Fix several warnings (#6242)

* Fix several warnings

Make class/struct definitions consistent( forward declararation with different type from definition )
Remove pessimising moves of temporary objects
Change `info` symbol in library_info.cc into non-static

Signed-off-by: Vladimir Plazun <v.plazun@samsung.com>
5 years agoReplace strict build flag setting in libs/tflite (#6232)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 03:40:39 +0000 (12:40 +0900)]
Replace strict build flag setting in libs/tflite (#6232)

Replace libs/tflite strict build setting cmake from using compile option to linking nnfw_common interface

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[res] TF Shape-Reshape network (#6343)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 02:45:27 +0000 (11:45 +0900)]
[res] TF Shape-Reshape network (#6343)

This commit adds tensorflow Shape test using Reshape node.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years agoReplace strict build setting in libs (#6226)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 02:44:47 +0000 (11:44 +0900)]
Replace strict build setting in libs (#6226)

Replace strict build setting in libs to use nnfw_common interface (except libs/tflite)

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[loco] Infer the shape of bias-taking nodes (#6332)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 02:41:16 +0000 (11:41 +0900)]
[loco] Infer the shape of bias-taking nodes (#6332)

* [loco] Infer the shape of bias-taking nodes

CanonicalShapeInferenceRule is now able to infer the shape of
BiasEncode, FeatureBiasAdd, and TensorBiasAdd nodes.

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

5 years ago[neurun] Enhance DotDumper backend color assign (#6368)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 02:27:18 +0000 (11:27 +0900)]
[neurun] Enhance DotDumper backend color assign (#6368)

Enhance DotDumper backend color assignment. Each Backend will have its
unique color up to 8 of them. Before it was assigned based on hard-coded
Backend ID.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[moco-tf] Introduce TFShape IR (#6322)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 01:52:17 +0000 (10:52 +0900)]
[moco-tf] Introduce TFShape IR (#6322)

This commit introduces TF-dialect TFShape IR for tensorflow 'Shape'
operation

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[loco] Shape inference over Conv2D-related nodes (#6349)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 01:47:06 +0000 (10:47 +0900)]
[loco] Shape inference over Conv2D-related nodes (#6349)

CanonicalShapeInferenceRule is now able to infer the shape of nodes
related with Conv2D (such as FilterEncode).

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] Resolve negative squeeze dimension (#6352)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 01:46:41 +0000 (10:46 +0900)]
[moco-tf] Resolve negative squeeze dimension (#6352)

This commit resolves negative squeeze dimension during shape inference

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[neurun] Favor `map::emplace` over `map::insert` (#6355)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 01:46:33 +0000 (10:46 +0900)]
[neurun] Favor `map::emplace` over `map::insert` (#6355)

Favor `map::emplace` over `map::insert` for every occurences in neurun.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[moco-tf] relocate AvgPool local class into anonymous (#6366)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 01:46:11 +0000 (10:46 +0900)]
[moco-tf] relocate AvgPool local class into anonymous (#6366)

This will relocate local class of AvgPool import into anonymous namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] relocate Add local class into anonymous (#6365)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 01:45:50 +0000 (10:45 +0900)]
[moco-tf] relocate Add local class into anonymous (#6365)

This will relocate local class of Add import into anonymous namespace

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Disable usage of nodes.lst in FixPadding (#6361)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 01:45:17 +0000 (10:45 +0900)]
[moco-tf] Disable usage of nodes.lst in FixPadding (#6361)

This will disable using CanonicalNodes.lst and TFNodes.lst so that we don't have to modify FixPaddingTransform
- this is a temporary fix before deleting FixPaddingTransform related files

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[neurun] Refine DotDumper (#6347)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 01:44:01 +0000 (10:44 +0900)]
[neurun] Refine DotDumper (#6347)

- Rename method `dumpIfNeeded` to `dump`
- Rename enum `OPTION` to `DotDumper::Level` and update comments
- Take out config fetching from class `DotDumper`

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[moco-tf] Import as TFConcatV2 with a knob (#6348)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 8 Aug 2019 01:42:30 +0000 (10:42 +0900)]
[moco-tf] Import as TFConcatV2 with a knob (#6348)

This will change import of ConcatV2 node as Canonical or TF-dialect with a knob.

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[loco] Hide deprecated "node" method (#6342)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 01:20:17 +0000 (10:20 +0900)]
[loco] Hide deprecated "node" method (#6342)

Now, "node" method in GraphInput/GraphOutput is private.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[res] Simple TF Shape test case (#6341)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 01:19:57 +0000 (10:19 +0900)]
[res] Simple TF Shape test case (#6341)

This commit adds simple Shape network for TF test.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years agoFix converting type of zero point in select_operator.py (#6353)
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 00:49:33 +0000 (09:49 +0900)]
Fix converting type of zero point in select_operator.py (#6353)

This commit fixes converting type of zero point in select_operator.py.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years agoRevise acl common cmake (#6340)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 00:45:46 +0000 (09:45 +0900)]
Revise acl common cmake (#6340)

- Add strict build option: nnfw_common
- Link arm_compute and arm_compute_ex library
- Include path acl_common for acl_common user (acl_cl, acl_neon)

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoMerge neurun kernel and backend build (#6313)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Aug 2019 00:45:27 +0000 (09:45 +0900)]
Merge neurun kernel and backend build (#6313)

- Merge neurun each backend's kernel and backend build
- Fix build error on cpu backend kernel by restrict build option
  - Invalid type comparison
  - Unused variable
  - Unused function

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[locop] FixedReshape node summary (#6354)
박천교/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 8 Aug 2019 00:22:43 +0000 (09:22 +0900)]
[locop] FixedReshape node summary (#6354)

This commit adds FixedReshape node summary.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
5 years ago[moco-tf] Use common calc_paddata in FixShape (#6346)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 22:01:57 +0000 (07:01 +0900)]
[moco-tf] Use common calc_paddata in FixShape (#6346)

This will change calc_node_pad() and calc_annot_paddata() to use common calc_annot_paddata() in FixShapeTransform

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[nnc] Replace BiasAdd and Scale with elementwise equivalents (#6358)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Wed, 7 Aug 2019 15:08:48 +0000 (18:08 +0300)]
[nnc] Replace BiasAdd and Scale with elementwise equivalents (#6358)

`BiasAdd` and `Scale` are restricted versions of equivalent Elementwise ops and are going to be removed.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years agoRemove TensorBuilder getter from IKernelGenerator (#6331)
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 7 Aug 2019 12:23:12 +0000 (21:23 +0900)]
Remove TensorBuilder getter from IKernelGenerator (#6331)

Now that `Backend` holds `TensorBuilder` directly, no need to have this
interface from `IkernelGenerator` any longer.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[mir] Add binary elementwise operations (#6351)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Wed, 7 Aug 2019 12:04:29 +0000 (15:04 +0300)]
[mir] Add binary elementwise operations (#6351)

first step of future replacement of single `ElementwiseOp` class with set of individual operations.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[moco-tf] Introduce TFConcatV2 IR (#6321)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 09:56:12 +0000 (18:56 +0900)]
[moco-tf] Introduce TFConcatV2 IR (#6321)

This will introduce TFConcatV2 IR and required codes for TensorFlow Concat node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years agoSupport asynchronous execution (#6314)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 09:44:42 +0000 (18:44 +0900)]
Support asynchronous execution (#6314)

- Introduce methods for start execution & wait finish in Execution class to support asynchronous execution
- Add test using introduced methods

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[tf2tflite] copying customop.conf into cmake build/compiler/tf2tflite (#6329)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 09:21:33 +0000 (18:21 +0900)]
[tf2tflite] copying customop.conf into cmake build/compiler/tf2tflite (#6329)

* [tf2tflite] copying customop.conf into cmake build/compiler/tf2tflite

This makes cmake to copy customop.conf into cmake build/compiler/tf2tflite.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* indent, 'xxx_customop.conf' -> 'xxx.customop.conf'

5 years ago[exo-tflite] Implement FilterShape converter (#6334)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 09:21:13 +0000 (18:21 +0900)]
[exo-tflite] Implement FilterShape converter (#6334)

This commit implement a helper which converts FilterShape as exo-tflite
internal ShapeDescription.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] Remove redundant codes in FuseBinaryIntoPreceding (#6319)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 7 Aug 2019 09:14:16 +0000 (18:14 +0900)]
[moco-tf] Remove redundant codes in FuseBinaryIntoPreceding (#6319)

* [moco-tf] Remove redundant codes in FuseBinaryIntoPreceding

There are some redundant codes in `FuseBinaryIntoPreceding.cpp`
This commit will remove them.
- Exactly only one of variables(`xc` and `yc`) is `nullptr` because `is_only_one_valid` already checks.
- Then `addparam` and `mulparam` cannot be `nullptr`.
- After removing above condition, there can be only one case that `bchanged` can be returned, that is when `biasadd` is `nullptr`. However when `biasadd` is `nullptr`, bchanged is always `false`.

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

* add missing codes

* fix comments

* apply comments

5 years ago[moco-tf] Common method for calculation of PadData (#6338)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 09:10:30 +0000 (18:10 +0900)]
[moco-tf] Common method for calculation of PadData (#6338)

This will introduce common method for calculation of PadData that exist in calc_node_pad() and calc_annot_paddata() of FixShapeTransform

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[loco] Introduce GraphBuilder helper class (#6308)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 08:22:09 +0000 (17:22 +0900)]
[loco] Introduce GraphBuilder helper class (#6308)

* [loco] Introduce GraphBuilder helper class

This commit introduces GraphBuilder helper class which faciliates
testcase implementation.

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

5 years ago[loco] TensorConcat Shape Inference (#6315)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 08:18:58 +0000 (17:18 +0900)]
[loco] TensorConcat Shape Inference (#6315)

* [loco] TensorConcat Shape Inference

CanonicalShapeInferenceRule is now able to infer the shape of  TensorConcat
nodes.

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

5 years ago[moco-tf] Deprecate FixPaddingTransform (#6335)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 08:18:46 +0000 (17:18 +0900)]
[moco-tf] Deprecate FixPaddingTransform (#6335)

This will remove using FixPaddingTransform.

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years agoRename pacl build script (#6336)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 08:11:06 +0000 (17:11 +0900)]
Rename pacl build script (#6336)

Rename pacl build script: suffix pacl
Rename to avoid default runtime confusion
Update build introduction document

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[tf2tflite] Modified as CustomOp.info renamed to customop.conf (#6328)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 07:53:50 +0000 (16:53 +0900)]
[tf2tflite] Modified as CustomOp.info renamed to customop.conf (#6328)

CMakeLists.txt file was modified as CustomOp.info renamed to customop.conf.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[moco-tf] Fix pad for TF dialect nodes in FixShape (#6302)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 06:42:17 +0000 (15:42 +0900)]
[moco-tf] Fix pad for TF dialect nodes in FixShape (#6302)

This will merge fix pad for TF dialect nodes in FixShapeInference using calc_annot_paddata() method

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Use calc_annot_paddata for fix pad of TF dialect nods (#6301)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 06:41:48 +0000 (15:41 +0900)]
[moco-tf] Use calc_annot_paddata for fix pad of TF dialect nods (#6301)

This will change to use calc_annot_paddata() method to fix pad in FixShapeInference for TFAvgPool node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[tf2tflite] driver can takes customop.conf as param (#6272)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 06:39:22 +0000 (15:39 +0900)]
[tf2tflite] driver can takes customop.conf as param (#6272)

* [tf2tflite] driver can takes customop.conf as param

This commit enable Driver to take customop.conf as the 5th param.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* optional param: '--customop <customop.conf path>'

* remove strcmp, parse conf before calling import

* nyi error

5 years ago[exo-tflite] Exporting Operator Code table including code of custom op (#6323)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 06:38:31 +0000 (15:38 +0900)]
[exo-tflite] Exporting Operator Code table including code of custom op (#6323)

Now when tflite opcode table is created, name and code for custom op is included.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[exo-tflite] type inferencing including custom op (#6320)
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 05:52:28 +0000 (14:52 +0900)]
[exo-tflite] type inferencing including custom op (#6320)

This commit enables exo-tflite to infer type canonical nodes as well as custop op.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[tf2tflite] Enable test RealDiv and Sub (#6325)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 05:41:48 +0000 (14:41 +0900)]
[tf2tflite] Enable test RealDiv and Sub (#6325)

This will enable tf2tflite test of RealDiv and Sub

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[mocotest-tf] Enable RealDiv and Sub test (#6324)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 05:40:43 +0000 (14:40 +0900)]
[mocotest-tf] Enable RealDiv and Sub test (#6324)

This will enable mocotest-tf test for RealDiv and Sub

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[loco] Identity Graph Testcase with Input Shape (#6307)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 05:40:25 +0000 (14:40 +0900)]
[loco] Identity Graph Testcase with Input Shape (#6307)

This commit allows users to build an identity graph testcase with
explicit input shape.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[moco-tf] Enable Canonicalization of Sub and RealDiv (#6316)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 05:16:27 +0000 (14:16 +0900)]
[moco-tf] Enable Canonicalization of Sub and RealDiv (#6316)

This will enable Canonicalizaton of TFSub and TFRealDiv nodes

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Canonicalizer for RealDiv (#6306)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 04:15:57 +0000 (13:15 +0900)]
[moco-tf] Canonicalizer for RealDiv (#6306)

This will introduce Canonicalizer for TFRealDiv to EltwiseDiv

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[mir-tflite-importer] Perform operation sanity checks while converting it (#6069)
Ivan Vagin/AI Tools Lab /SRR/Engineer/삼성전자 [Wed, 7 Aug 2019 04:09:53 +0000 (07:09 +0300)]
[mir-tflite-importer] Perform operation sanity checks while converting it (#6069)

* [mir-tflite-importer] Perform operation sanity checks while creating

Perform operation sanity checks while creating it

Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
5 years agoRemove format check warning by deleted file (#6312)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 03:45:29 +0000 (12:45 +0900)]
Remove format check warning by deleted file (#6312)

By filter out deleted file, remove warning on format checker's diff with master mode

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoEnable DETECTION_POSTPROCESS kernel in TFLite interpreter (#6214)
이춘석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 02:26:31 +0000 (11:26 +0900)]
Enable DETECTION_POSTPROCESS kernel in TFLite interpreter (#6214)

- Enable DETECTION_POSTPROCESS in interpreter

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
5 years ago[moco-tf] Canonicalizer for Sub (#6305)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 01:55:30 +0000 (10:55 +0900)]
[moco-tf] Canonicalizer for Sub (#6305)

This will introduce Canonicalizer for TFSub node to EltwiseSub

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Graph build for RealDiv node (#6304)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 01:54:52 +0000 (10:54 +0900)]
[moco-tf] Graph build for RealDiv node (#6304)

This will enable graph building of RealDiv node for importer

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Graph build for Sub node (#6303)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 01:54:29 +0000 (10:54 +0900)]
[moco-tf] Graph build for Sub node (#6303)

This will enable graph building of Sub node for importer

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Use calc_node_pad for fix pad of loco nodes (#6300)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 01:42:35 +0000 (10:42 +0900)]
[moco-tf] Use calc_node_pad for fix pad of loco nodes (#6300)

This will change to use calc_node_pad() method to fix pad in FixShapeInference for loco nodes

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[moco-tf] Fix pad for TFConv2D in fix shape (#6250)
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 01:42:10 +0000 (10:42 +0900)]
[moco-tf] Fix pad for TFConv2D in fix shape (#6250)

This will merge fix pad for TFConv2D node in fix shape transform

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
5 years ago[loco] Support ConstGen shape inference (#6290)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 01:31:28 +0000 (10:31 +0900)]
[loco] Support ConstGen shape inference (#6290)

CanonicalShapeInferenceRule is now able to infer the shape of ConstGen
nodes.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[exo-tflite] BiasShape to ShapeDescription Converter (#6289)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 01:30:01 +0000 (10:30 +0900)]
[exo-tflite] BiasShape to ShapeDescription Converter (#6289)

This commit implements BiasShape -> ShapeDescription converter.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years agoIntroduce "xtrace" (#6158)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 01:16:30 +0000 (10:16 +0900)]
Introduce "xtrace" (#6158)

* Introduce "xtrace"

This commit introduces "xtract" which collects trace/benchmark events
via xray pipe and records it as Chrome Trace Event file format.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Add missing copyright

5 years ago[loco] MaxPool2D Shape Inference (#6287)
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 00:44:41 +0000 (09:44 +0900)]
[loco] MaxPool2D Shape Inference (#6287)

CanonicalShapeInferenceRule is now able to infer the shape of MaxPool2D
nodes.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
5 years ago[nnpackage] Add an nnpackage example (one op, tflite) (#5929)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 00:43:24 +0000 (09:43 +0900)]
[nnpackage] Add an nnpackage example (one op, tflite) (#5929)

Add first minimal nnpackage example.
It has no custom op, one operator, uses tflite model.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
5 years ago[nnpackage] nnpackager -c will create compressed nnpackage (#6278)
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 7 Aug 2019 00:42:59 +0000 (09:42 +0900)]
[nnpackage] nnpackager -c will create compressed nnpackage (#6278)

Use "nnpackage -c path_to_nnpackage_root" to get compressed nnpackage.
Also it removed redundant parenthesis in if condition.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
5 years agoReplace strict build flag setting in neurun frontend & core (#6279)
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Aug 2019 00:40:54 +0000 (09:40 +0900)]
Replace strict build flag setting in neurun frontend & core (#6279)

- Replace neurun frontends and core strict build setting cmake from using compile option to linking nnfw_common interface
- Add strict build setting for nnfw api frontend and fix unused parameter warning

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[moco-tf] Introduce function to create BiasAdd operation (#6292)
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 7 Aug 2019 00:03:26 +0000 (09:03 +0900)]
[moco-tf] Introduce function to create BiasAdd operation (#6292)

This commit will introduce a new function to create `BiasAdd` operation
and insert it after preceding operation.

Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
5 years agoAdd default fallback for Visitor::visit (#5889)
Дилшоджон Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Tue, 6 Aug 2019 19:49:39 +0000 (22:49 +0300)]
Add default fallback for Visitor::visit (#5889)

This will help to escape overriding a lot of visit()
with default body: empty body or `throw std::runtime_error("NYI")`:
just override this new method

additional changes:
* added file with all mir operation headers
* refined IVisitor

Signed-off-by: Dilshodzhon Poshshoev <d.poshshoev@samsung.com>
5 years ago[mir_onnx] Replace BiasAdd and Scale with Elementwise equivalents (#6298)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Tue, 6 Aug 2019 18:49:35 +0000 (21:49 +0300)]
[mir_onnx] Replace BiasAdd and Scale with Elementwise equivalents (#6298)

`BiasAdd` and `Scale` are restricted versions of equivalent Elementwise ops and are going to be removed.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
5 years ago[mir2loco] Replace BiasAdd and Scale with Elementwise equivalents (#6296)
Сергей Баранников/AI Tools Lab /SRR/Engineer/삼성전자 [Tue, 6 Aug 2019 17:07:39 +0000 (20:07 +0300)]
[mir2loco] Replace BiasAdd and Scale with Elementwise equivalents (#6296)

`BiasAdd` and `Scale` are restricted versions of equivalent Elementwise ops and are going to be removed.

Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>