platform/core/ml/nnfw.git
5 years agoIntroduce Compiler class (#3632)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 21 Nov 2018 01:41:19 +0000 (10:41 +0900)]
Introduce Compiler class (#3632)

Introduce neurun::compiler::Compiler class to hide compilation phase from frontend
Introduce src/compiler directory for compile phases

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoAdded generated test data for CUSTOM unpack ops (#3638)
Vishal Keshav/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 21 Nov 2018 01:26:50 +0000 (06:56 +0530)]
Added generated test data for CUSTOM unpack ops (#3638)

The input tensor data is treated as buffer.
Along any given axis, the output buffer are to be generated
according to the strides (determined by axis and num_splits)

Signed-off-by: Vishal Keshav <vishal.k1@samsung.com>
5 years agoIntroduce LOGICAL_OR_EX in NeuralNetworksEx.h (#3618)
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 21 Nov 2018 01:01:17 +0000 (06:31 +0530)]
Introduce LOGICAL_OR_EX in NeuralNetworksEx.h (#3618)

This patch Introduces LOGICAL_AND_EX in NeuralNetworksEx.h.
Related issue: #3459

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoGenerate text file to save build path depth in coverage build (#3651)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 20 Nov 2018 10:09:36 +0000 (19:09 +0900)]
Generate text file to save build path depth in coverage build (#3651)

Generate text file tools/test_driver/build_path_depth.txt to save build path depth.
gcov for cross build need directory strip to match build host directory and target directory.
Build path depth will be used to set GCOV_PREFIX_STRIP environment variable on target.

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoFix comment for argmax_ex in NeuralNetworksEx.h (#3606)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 20 Nov 2018 10:09:14 +0000 (19:09 +0900)]
Fix comment for argmax_ex in NeuralNetworksEx.h (#3606)

Fix incorrect comment for output of argmax_ex in NeuralNetworksEx.h
- Output rank should be same rank with input
- Reduced axes are retained with length 1

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoNeuralNetworkEx code cleanup (#3649)
TANUJ TEKRIWAL/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 07:24:30 +0000 (12:54 +0530)]
NeuralNetworkEx code cleanup (#3649)

Just little code cleanup for NeuralNetworkEx.h

Signed-off-by: Tanuj Tekriwal <tanuj.tekri@samsung.com>
5 years ago[neurun] Fix TensorBuilder to add subtensors to Plan::operands (#3614)
Дилшоджон Умронхонович Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Tue, 20 Nov 2018 07:19:00 +0000 (10:19 +0300)]
[neurun] Fix TensorBuilder to add subtensors to Plan::operands (#3614)

Fix TensorBuilder to add subtensors to Plan::operands
Related issue is #3611

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
5 years agoIntroduce LOGICAL_AND_EX in NeuralNetworksEx.h (#3563)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 07:16:52 +0000 (12:46 +0530)]
Introduce LOGICAL_AND_EX in NeuralNetworksEx.h (#3563)

This patch Introduces LOGICAL_AND_EX in NeuralNetworksEx.h.
Related issue: #3459

Signed-off-by: prasannar <prasanna.r@samsung.com>
5 years agoAdded Generic test file for RSQRT_EX (#3297)
TANUJ TEKRIWAL/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 05:25:46 +0000 (10:55 +0530)]
Added Generic test file for RSQRT_EX (#3297)

This patch adds the Gtest file for RSQRT_EX.

Signed-off-by: Tanuj Tekriwal <tanuj.tekri@samsung.com>
5 years agoPass all arguments in summarize_tflite tool (#3646)
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 20 Nov 2018 05:13:25 +0000 (14:13 +0900)]
Pass all arguments in summarize_tflite tool (#3646)

To use various options of model_parser.py tool,
it allows to pass all arguments to model_parser.py file in summarize_tflite file.

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
5 years agoFix non-string passed to CreateString issue in tfltool (#3645)
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 20 Nov 2018 05:13:15 +0000 (14:13 +0900)]
Fix non-string passed to CreateString issue in tfltool (#3645)

If the operator_codes is empty, select_operator.py tool occurs an error.
This patch is fixed the issue.

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
5 years ago[neurun] Rename Planner as OperationValidator (#3582)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 20 Nov 2018 05:01:14 +0000 (14:01 +0900)]
[neurun] Rename Planner as OperationValidator (#3582)

Now Planner has only have assertions, so rename accordingly.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoAdd gtest to test for quant8 input for Pad op (#3620)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 04:17:25 +0000 (09:47 +0530)]
Add gtest to test for quant8 input for Pad op (#3620)

This patch will add gtest to test for quant8 input for pad op.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoAdd GTests for SQRT op (#3608)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 04:17:12 +0000 (09:47 +0530)]
Add GTests for SQRT op (#3608)

This patch addds GTests for SQRT op.
Added files:
```
sqrt_ex_1D_float
sqrt_ex_4D_float
```

Signed-off-by: prasannar <prasanna.r@samsung.com>
5 years agoRemoving unnecessary included header files in NENormalization (#3621)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 04:06:13 +0000 (09:36 +0530)]
Removing unnecessary included header files in NENormalization (#3621)

This patch will remove unnecessary headers included in src and runtime
of NENormalization.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoIntroduce LogicalOr op in runtime (#3622)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 04:05:59 +0000 (09:35 +0530)]
Introduce LogicalOr op in runtime (#3622)

This patch will add node for LogicalOr op in PACL

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoCorrecting the comments in equal and notequal cl (#3627)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 04:05:41 +0000 (09:35 +0530)]
Correcting the comments in equal and notequal cl (#3627)

This patch will corrects the comments in equal and notequal cl
And Adds the missing comments.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoRemoving unnecessary headers included in clfiles of Permute op (#3628)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 04:05:26 +0000 (09:35 +0530)]
Removing unnecessary headers included in clfiles of Permute op (#3628)

This patch will remove unnecessary headers included in clfiles of Permute op.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoCorrection in Error checking conditions and namespace missalignment (#3630)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 04:05:14 +0000 (09:35 +0530)]
Correction in Error checking conditions and namespace missalignment (#3630)

This patch will correct:
1. error checking condition for output, since output is always QASSYM8
input1 and output datatypes cannot be compared. Also, output can omnly be
checked for QASSYM8 in validate function.

2.Since all the non class functions are placed in namespace (format followed
by ACL) hence placing validate the function in the defined namespace.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoMinor changes in TransposeConv op (#3619)
Devansh Bansal/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 20 Nov 2018 04:04:45 +0000 (09:34 +0530)]
Minor changes in TransposeConv op (#3619)

* Minor changes in TransposeConv op

Addressing the comments by jh1302-park in #3423

Signed-off-by: b.devansh <b.devansh@samsung.com>
* Minor changes in TransposeConv op

Addressing the comments by jh1302-park in #3423

Signed-off-by: b.devansh <b.devansh@samsung.com>
5 years agoIntroducing LogicalNot op in NeuralNetworksEx (#3598)
Дилшоджон Умронхонович Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Tue, 20 Nov 2018 04:03:30 +0000 (07:03 +0300)]
Introducing LogicalNot op in NeuralNetworksEx (#3598)

This commit introduces LogicalNot op in NeuralNetworksEx.h
Related issues: #3596

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
5 years ago[model_parser] Fix wrong printing type of tensor (#3635)
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 20 Nov 2018 02:03:54 +0000 (11:03 +0900)]
[model_parser] Fix wrong printing type of tensor (#3635)

* [model_parser] Fix wrong printing type of tensor

Fixes wrong printing type of tensor. The type names like FLOAT32,
UINT8 not 0 should be printed.

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
* Fix to get type_name from tensor_wrapping

5 years agoNnapi delegate for ArgMax (#3558)
PRAVEEN DORESWAMY NAIDU/System SW /SRI-Bangalore/Staff Engineer/삼성전자 [Mon, 19 Nov 2018 08:21:52 +0000 (13:51 +0530)]
Nnapi delegate for ArgMax (#3558)

The changes introduces the nnapi delegate for ArgMax operation.

Signed-off-by: Praveen D N <praveen.dn@samsung.com>
5 years agoAdd generated test for ARG_MAX (#3554)
PRAVEEN DORESWAMY NAIDU/System SW /SRI-Bangalore/Staff Engineer/삼성전자 [Mon, 19 Nov 2018 07:46:27 +0000 (13:16 +0530)]
Add generated test for ARG_MAX (#3554)

The changes adds the generated test's for ARG max operation.

Signed-off-by: Praveen D N <praveen.dn@samsung.com>
5 years agoApply CL Kernel of Embedding_Lookup to PACL (#3616)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 19 Nov 2018 07:44:12 +0000 (16:44 +0900)]
Apply CL Kernel of Embedding_Lookup to PACL (#3616)

This commit applies CL Kernel of Embedding_Lookup to PACL.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years agoAdd gtest to test Transpose op for Tensors with rank < 4 (#3601)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Mon, 19 Nov 2018 02:35:03 +0000 (08:05 +0530)]
Add gtest to test Transpose op for Tensors with rank < 4 (#3601)

This patch will add gtest to verify Transpose op for input tensors
having rank < 4

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years ago[model_parser] Prints optons for Depthwise_Conv_2D (#3613)
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 19 Nov 2018 02:02:46 +0000 (11:02 +0900)]
[model_parser] Prints optons for Depthwise_Conv_2D (#3613)

This commit adds feature to print optons for Depthwise_Conv_2D.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years ago[model parser] Fix parsing bug when operator don't have option (#3610)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 19 Nov 2018 02:02:35 +0000 (11:02 +0900)]
[model parser] Fix parsing bug when operator don't have option (#3610)

Fix parsing bug invaild exit when operator don't have option

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoChange SimpleSpaceToBatchND's inputs to arm_compute::ITensor (#3553)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 19 Nov 2018 01:22:51 +0000 (10:22 +0900)]
Change SimpleSpaceToBatchND's inputs to arm_compute::ITensor (#3553)

This commit changes SimpleSpaceToBatchND's inputs to arm_compute::ITensor.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years agoAdd freeze_graph script into tflkit (#3556)
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 16 Nov 2018 08:13:41 +0000 (17:13 +0900)]
Add freeze_graph script into tflkit (#3556)

freeze_graph tool will freeze a standalone graphdef with the stored variables.

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
5 years agoAdd GTest for testing the enabled Broadcast of EqualEx op (#3304)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Fri, 16 Nov 2018 07:21:13 +0000 (12:51 +0530)]
Add GTest for testing the enabled Broadcast of EqualEx op (#3304)

This patch adds GTest for testing the enabled Broadcast of EqualEx op.
Added GTests:
```
equal_ex_broadcast
equal_ex_broadcast_4D_2D
```
Signed-off-by: prasannar <prasanna.r@samsung.com>
5 years agoImplementation of TransposeConv op in PACL for CPU (#3423)
Devansh Bansal/System SW /SRI-Bangalore/Engineer/삼성전자 [Fri, 16 Nov 2018 07:17:30 +0000 (12:47 +0530)]
Implementation of TransposeConv op in PACL for CPU (#3423)

This patch will add TransposeConv op
Related issues : #2739 #2809

Signed-off-by: b.devansh <b.devansh@samsung.com>
5 years agoImplement CL kernel of EmbeddingLookup (#3489)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 16 Nov 2018 06:30:26 +0000 (15:30 +0900)]
Implement CL kernel of EmbeddingLookup (#3489)

This commit implements CL kernel of EmbeddingLookup op.
  - Add an IFunction class for EmbeddingLookup.
  - Add an ICLKernel class for EmbeddingLookup.
  - Add cl kernel functions for EmbeddingLookup.
  - Append the cl functions to kernel_program_map.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years agoIntroducing LogicalNot op in pure_arm_compute runtime (#3599)
Дилшоджон Умронхонович Пошшоев/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 16 Nov 2018 04:24:37 +0000 (07:24 +0300)]
Introducing LogicalNot op in pure_arm_compute runtime (#3599)

This commit introduces LogicalNot op in in pure_arm_compute runtime
Related PR is #3598
Related issues: #3596 and #3459

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
5 years agoPack generated test update (#3557)
PRAVEEN DORESWAMY NAIDU/System SW /SRI-Bangalore/Staff Engineer/삼성전자 [Fri, 16 Nov 2018 04:20:23 +0000 (09:50 +0530)]
Pack generated test update (#3557)

The changes updates the pack operation in all_generated_test.

Signed-off-by: Praveen D N <praveen.dn@samsung.com>
5 years agoRevise x86 build script (#3604)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 16 Nov 2018 02:43:23 +0000 (11:43 +0900)]
Revise x86 build script (#3604)

x86 build by non-root anomymous user: same userid with host user
Use docker-run-user command and remove docker setting code

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoAdd transform_graph script into tflkit (#3545)
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 15 Nov 2018 07:27:31 +0000 (16:27 +0900)]
Add transform_graph script into tflkit (#3545)

* Add transform_graph script into tflkit

transform_graph tool will transform a TensorFlow GraphDef applying some transformations.

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Update detailed usage to README.md

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
5 years ago[tfltool] Update option parsing in select operator for v1.12 (#3594)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 15 Nov 2018 07:10:18 +0000 (16:10 +0900)]
[tfltool] Update option parsing in select operator for v1.12 (#3594)

Add comment for new options
Fix DepthwiseConv2DOptions for additional field

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Remove IPlanBuilder::addShapeConstr (#3591)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 15 Nov 2018 04:32:30 +0000 (13:32 +0900)]
[neurun] Remove IPlanBuilder::addShapeConstr (#3591)

Since #2741 is merged, we do not need this method. We can remove this
from IPlanBuilder.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[neurun] DotDumper shows model outputs (#3590)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 15 Nov 2018 04:17:01 +0000 (13:17 +0900)]
[neurun] DotDumper shows model outputs (#3590)

Just like model inputs, this commit shows model outputs as a different
shape.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[tfltool] Update tflite schema API version (#3581)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 15 Nov 2018 04:14:16 +0000 (13:14 +0900)]
[tfltool] Update tflite schema API version (#3581)

Update tflite schema API version v1.9 -> v1.12

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoSupport docker run as non-root user (#3592)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 15 Nov 2018 04:03:19 +0000 (13:03 +0900)]
Support docker run as non-root user (#3592)

Run non-root anomymous user: same userid with host user
Used on arm cross build and tizen build

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Introduce Model class (#3586)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 15 Nov 2018 02:41:50 +0000 (11:41 +0900)]
[neurun] Introduce Model class (#3586)

Model class is a container of actual operands/operations and model's
inputs/outputs.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[model_parser] Supports printing specific tensors and operators (#3580)
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 15 Nov 2018 02:09:24 +0000 (11:09 +0900)]
[model_parser] Supports printing specific tensors and operators (#3580)

Now model_parser.py supports argument options for features which prints
specific tensors and operators.

./model_parser.py ./tflite -t 1 2 3   # tensor id
./model_parser.py ./tflite -o 1 2 3   # operator id

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years agoImplementing NotEqual op in runtime (#3515)
Siva Sai Vaddipati/System SW /SRI-Bangalore/Engineer/삼성전자 [Thu, 15 Nov 2018 01:10:01 +0000 (06:40 +0530)]
Implementing NotEqual op in runtime (#3515)

This commit implements the CL version of NotEqual op.
Related issues: #3335, #3337

Signed-off-by: Siva Sai <siva.sai@samsung.com>
5 years agoRemoving TODO for mul (#3363)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Wed, 14 Nov 2018 23:36:31 +0000 (05:06 +0530)]
Removing TODO for mul (#3363)

This patch will remove TODO for mul op.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoAdd Pad gtests to test padding across any dimention (#3308)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Wed, 14 Nov 2018 23:36:14 +0000 (05:06 +0530)]
Add Pad gtests to test padding across any dimention (#3308)

This patch will add gtests to test padding across any dimension.
This will verify padding op is working when in addition to spatial padding,
Padding is done across batch or depth too.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoAdd comment for argmax_ex (#3576)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 14 Nov 2018 09:02:43 +0000 (18:02 +0900)]
Add comment for argmax_ex (#3576)

Add comment for ANEURALNETWORKS_ARGMAX_EX in NeuralNetworksEx.h

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Remove NDEBUG in planTensors (#3579)
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 14 Nov 2018 08:21:56 +0000 (17:21 +0900)]
[neurun] Remove NDEBUG in planTensors (#3579)

* [neurun] Remove NDEBUG in planTensors

Removes NDEBUG code in Linear::planTensors.

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

5 years agoDefine TENSORFLOW_SUM in NeuralNetworksEx.h (#3578)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Wed, 14 Nov 2018 07:35:04 +0000 (13:05 +0530)]
Define TENSORFLOW_SUM in NeuralNetworksEx.h (#3578)

This patch defines TENSORFLOW_SUM in NeuralNetworksEx.h.
Related issue: #3347.

Signed-off-by: prasannar <prasanna.r@samsung.com>
5 years ago[neurun] Refine DAGChecker implementation (#3538)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 14 Nov 2018 07:22:05 +0000 (16:22 +0900)]
[neurun] Refine DAGChecker implementation (#3538)

Make DAGChecker implementation simpler by using operands' use info.
Cannot make a cyclic test for now, since after creating a cycle then
`finishBuilding` is called, it will cause assert failure by runtime.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoAdd gtest for mul op to test 4D tensors (#3452)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Wed, 14 Nov 2018 04:47:09 +0000 (10:17 +0530)]
Add gtest for mul op to test 4D tensors (#3452)

This patch will add gtest to test mul op for 4D tensors

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years ago[model_parser] Print options of Conv2D, AvgPool2D, MaxPool2D (#3573)
윤현식/동작제어Lab(SR)/Principal Engineer/삼성전자 [Wed, 14 Nov 2018 04:34:31 +0000 (13:34 +0900)]
[model_parser] Print options of Conv2D, AvgPool2D, MaxPool2D (#3573)

Through this commit, model parser prints options of Conv2D, AvgPool2D, MaxPool2D such as stride, padding, etc.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
5 years agoAdd missing comments in NodeVisitor.h (#3575)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Wed, 14 Nov 2018 04:23:15 +0000 (09:53 +0530)]
Add missing comments in NodeVisitor.h (#3575)

This patch adds the missing comments in NodeVisitor.h

Signed-off-by: prasannar <prasanna.r@samsung.com>
5 years agoImplement SQRT op in PACL (#3418)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Wed, 14 Nov 2018 01:27:24 +0000 (06:57 +0530)]
Implement SQRT op in PACL (#3418)

This patch implements SQRT op in PACL.
Related issue: #3341

Signed-off-by: prasannar <prasanna.r@samsung.com>
5 years agoAdd optimize_for_inference script into tflkit (#3546)
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 14 Nov 2018 00:46:35 +0000 (09:46 +0900)]
Add optimize_for_inference script into tflkit (#3546)

* Add optimize_for_inference script into tflkit

optimize_for_inference tool optimizes a TensorFlow GraphDef for inference.

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Add detail information to README.md

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
5 years agoAdd NNAPI delegation for SQRT_EX (#3422)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 13 Nov 2018 09:43:02 +0000 (15:13 +0530)]
Add NNAPI delegation for SQRT_EX (#3422)

This patch adds NNAPI delegation for SQRT_EX

Signed-off-by: prasannar <prasanna.r@samsung.com>
5 years ago[neurun] Remove Planner implementation (#3560)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 13 Nov 2018 09:34:26 +0000 (18:34 +0900)]
[neurun] Remove Planner implementation (#3560)

Now that we are able to have uniform planning procedure for all Nodes,
We do not need Planner class anymore. The planning procedure is now
unified and moved to `compilation.cc` but Planner class is left for
assertion checks which will be renamed as `OperationAsserter` later.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoMismatch in generated test files for NOT_EQUAL_EX (#3550)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 13 Nov 2018 07:55:14 +0000 (13:25 +0530)]
Mismatch in generated test files for NOT_EQUAL_EX (#3550)

This patches the mismatch in the GTest files for NOT_EQUAL_EX.

Signed-off-by: prasannar <prasanna.r@samsung.com>
5 years agoAdd tflite_convert scripts into tflkit (#3544)
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 13 Nov 2018 01:51:24 +0000 (10:51 +0900)]
Add tflite_convert scripts into tflkit (#3544)

* Add tflite_convert scripts into tflkit

tflite_convert tool will convert a TensorFlow GraphDef into TfLite model.

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Add check routine for bazel

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Change the PB_PATH to GRAPHDEF_PATH and add detail usages

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Add detail usage to README.md

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Fixed wrong exitcode

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
5 years ago[neurun] Add const iterate method to Linear (#3547)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 13 Nov 2018 01:51:04 +0000 (10:51 +0900)]
[neurun] Add const iterate method to Linear (#3547)

Add const iterate method to Linear

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years ago[neurun] Revise IStageGenerator (#3536)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 12 Nov 2018 04:12:55 +0000 (13:12 +0900)]
[neurun] Revise IStageGenerator (#3536)

* [neurun] Revise IStageGenerator

Revise IStageGenerator to be a NodeVisitor. This will make it possible
to unify Planner implementation.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
* Pass Stage as const reference

5 years agoUse pre-built acl by environment variable (#3533)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 12 Nov 2018 02:06:02 +0000 (11:06 +0900)]
Use pre-built acl by environment variable (#3533)

* Use pre-built acl by environment variable

When environment variable EXT_ACL_FOLDER is set, use pre-built acl binary

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Fix if-else syntax in Makefile

5 years agoAdding testcases for NotEqual Op (#3516)
Siva Sai Vaddipati/System SW /SRI-Bangalore/Engineer/삼성전자 [Fri, 9 Nov 2018 07:50:55 +0000 (13:20 +0530)]
Adding testcases for NotEqual Op (#3516)

This commit adds Gtest testcases for NotEqual op.
Related issues: #3335,#3337

Signed-off-by: Siva Sai <siva.sai@samsung.com>
5 years agoConfigure parallel execution job count (#3531)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 9 Nov 2018 06:50:12 +0000 (15:50 +0900)]
Configure parallel execution job count (#3531)

User can configure parallel execution job count using NPROCS environment variable
- Reduce local compilation overhead when doing multiple job
- Protect excessive CI processor consumtion by one job

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoCreate tflkit directory (#3438)
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 9 Nov 2018 05:20:39 +0000 (14:20 +0900)]
Create tflkit directory (#3438)

* Create tflite_easytool directory

This directory is created to make it easier to run frequently used tools in scripts.
Now, it has two functions:
- Summraize PB model
- Summraize TfLite model

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Change the directory name to tflkit

Also changes the relative path to absolute path

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Add Prerequires and TF tool guideline

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Add additional description for prerequires

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
5 years ago[neurun] Default implementation for NodeVisitor (#3530)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 9 Nov 2018 03:34:47 +0000 (12:34 +0900)]
[neurun] Default implementation for NodeVisitor (#3530)

Implement default implementation for NodeVisitor which is just doing
nothing.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoEnable concat elimination on acl backend (#3528)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 08:55:29 +0000 (17:55 +0900)]
Enable concat elimination on acl backend (#3528)

Enable concat elimination on acl backend
Change config and remove assertion block

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Remove IInitializerGenerator (#3525)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 8 Nov 2018 08:16:57 +0000 (17:16 +0900)]
[neurun] Remove IInitializerGenerator (#3525)

Remove IInitializerGenerator and its implementations. We do not need
this backend API anymore.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoAdd comment for tensorflow_max_ex (#3526)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 08:16:17 +0000 (17:16 +0900)]
Add comment for tensorflow_max_ex (#3526)

Add comment for tensorflow_max_ex in NeuralNetworksEx.h

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoFix typo in nnapi delegate (#3527)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 08:15:49 +0000 (17:15 +0900)]
Fix typo in nnapi delegate (#3527)

Fix typo in nnapi delegate: v1.11 -> v1.12

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Apply IMemoryPlanner to backend/cpu (#3517)
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 8 Nov 2018 07:48:44 +0000 (16:48 +0900)]
[neurun] Apply IMemoryPlanner to backend/cpu (#3517)

Now IMemoryPlanner and Allocator plays the role of IMemoryAllocator.
This commits includes some fixing for bugs to apply on backend/cpu.

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years agoUpdate tensorflow to v1.12 (#3486)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 07:12:55 +0000 (16:12 +0900)]
Update tensorflow to v1.12 (#3486)

* Update external package

- Update external package version
- Update cmake to build tflite
- Add gitignore for new external code

* Update library to support tflite

- Update nnapi delegate
- Update NeuralNetworksShim.h for new API
- Fix cmake for new files in updated tensorflow
- Workaround: Introduce OBS_BUILD flag to avoid gbs build fail

* Update neurun CPU kernel

- Update neurun CPU kernel to call updated tflite interpreter kernel
- Add build option -fPIC for updated tflite kernel

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoRegister subtensor info to tensor builder (#3519)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 06:49:08 +0000 (15:49 +0900)]
Register subtensor info to tensor builder (#3519)

Register subtensor info to tensor builder if
- Tensor is subtensor of other tensor
- Backend support subtensor allocation

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoCheck external directory exists (#3522)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 06:40:53 +0000 (15:40 +0900)]
Check external directory exists (#3522)

Check external directory exists and download if directory does not exist

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoFix run_benchmark.sh typo(MODLES->MODELS) (#3523)
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 05:12:20 +0000 (14:12 +0900)]
Fix run_benchmark.sh typo(MODLES->MODELS) (#3523)

- Fix run_benchmark.sh typo( MODLES -> MODELS )

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
5 years agoFix CI benchmark failure (#3520)
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 04:08:02 +0000 (13:08 +0900)]
Fix CI benchmark failure (#3520)

CI benchmark job fails since it does not reflect model path
change. This commit sets the proper model path.

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
5 years agoConcat elimination in codegen (#3518)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 02:44:39 +0000 (11:44 +0900)]
Concat elimination in codegen (#3518)

Concat elimination in codegen: check output tensor is parent of inputs and do nothing if true

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoBuild subtensor in acl TensorBuilder (#3513)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 01:56:03 +0000 (10:56 +0900)]
Build subtensor in acl TensorBuilder (#3513)

* Build subtensor in acl TensorBuilder

Build subtensor in acl TensorBuilder
Change to public function in Convert.cc

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Remove deprecated declaration

* Remove diff by space line

5 years ago[neurun] Remove InitializerGenerator dependency (#3508)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 8 Nov 2018 01:11:02 +0000 (10:11 +0900)]
[neurun] Remove InitializerGenerator dependency (#3508)

Remove all InitializerGenerator dependent code since we now have
ConstantInitializer for this purpose.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoIntroduce subtensor allocation check API (#3510)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 8 Nov 2018 01:10:32 +0000 (10:10 +0900)]
Introduce subtensor allocation check API (#3510)

Check API will be used in codegen

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Introduce MemoryPlanner and Allocator (#3511)
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 8 Nov 2018 01:09:12 +0000 (10:09 +0900)]
[neurun] Introduce MemoryPlanner and Allocator (#3511)

Introduces IMemoryPlanner, BumpPlanner, FirstFitPlanner and
Allocator to replace IMemoryAllocator.
Current IMemoryAllocator has two roles, Planner and Allocator. This
commits separates IMemoryAllocator to the two classes. After this commit
lands, IMemoryAllocator will be removed soon.

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years agoOptimize the StridedSlice Kernel for performance improvement (#3230)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 7 Nov 2018 09:31:46 +0000 (18:31 +0900)]
Optimize the StridedSlice Kernel for performance improvement (#3230)

This commit optimizes the StridedSlice Kernel for performance inprovement.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years agoIntroducing NotEqual Op in runtime (#3485)
Siva Sai Vaddipati/System SW /SRI-Bangalore/Engineer/삼성전자 [Wed, 7 Nov 2018 08:18:53 +0000 (13:48 +0530)]
Introducing NotEqual Op in runtime (#3485)

This commit introduces NotEqual op in runtime.
Related issues: #3335, #3337

Signed-off-by: Siva Sai <siva.sai@samsung.com>
5 years agoArgMinMax Implementation for ArgMin and ArgMax. (#3331)
PRAVEEN DORESWAMY NAIDU/System SW /SRI-Bangalore/Staff Engineer/삼성전자 [Wed, 7 Nov 2018 08:17:28 +0000 (13:47 +0530)]
ArgMinMax Implementation for ArgMin and ArgMax. (#3331)

This change introduces the cpu kernel for the ArgMin and ArgMax
operations.

Signed-off-by: Praveen D N <praveen.dn@samsung.com>
5 years ago[neurun] Fix build break (#3507)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 7 Nov 2018 08:01:26 +0000 (17:01 +0900)]
[neurun] Fix build break (#3507)

Caused by #3426 and #3436 collision.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoUpdate framework test list (#3499)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Nov 2018 07:40:33 +0000 (16:40 +0900)]
Update framework test list (#3499)

Add more framework test for neurun
Change inception test: use simple module instead of full models

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoIntroduce subtensor map (#3493)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Nov 2018 07:17:33 +0000 (16:17 +0900)]
Introduce subtensor map (#3493)

- Introduce subtensor map in ACL tensor builder
- TensorBuilder::at() and TensorBuilder::tensorAt() can return subtensor

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[neurun] Introduce ConstantInitializer (#3426)
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 7 Nov 2018 07:02:25 +0000 (16:02 +0900)]
[neurun] Introduce ConstantInitializer (#3426)

* [neurun] Introduce ConstantInitializer

This commit introduces `ConstantIntializer` class that fills operand
objects with the constant values which were specified by
`ANeuralNetworksModel_setOperandValue`. Plus, this class replaces the
InitializerGenerator.

NOTE due to FullyConnected operation's implicit reshaping, it is
implemented with workarounds.

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
* Add license

Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
5 years agoFix custom operator register macro (#3495)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Nov 2018 06:34:34 +0000 (15:34 +0900)]
Fix custom operator register macro (#3495)

Fix custom operator register macro for updated TfLiteRegistration field
- First 4 fields may not change
- Can adjust new fields before custom_name

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoEnable more model tests (#3497)
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Nov 2018 06:34:20 +0000 (15:34 +0900)]
Enable more model tests (#3497)

The following tests are triggered on default on framework/verification
test :

- embedding lookup
- L2 normalization
- L2 pool
- Squeeze
- Tanh
- Transpose
- Logistic

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
5 years agoIntroduce namespace in runtime load helper (#3494)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 7 Nov 2018 06:18:40 +0000 (15:18 +0900)]
Introduce namespace in runtime load helper (#3494)

Introduce namespace in runtime load helper
Restore function name
It protects naming conflict with upstream tensorflow

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years ago[model_parser] Hide instruction counts by verbose option (#3487)
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 7 Nov 2018 03:12:34 +0000 (12:12 +0900)]
[model_parser] Hide instruction counts by verbose option (#3487)

Hides instruction counts by verbose option. Now -v or --verbose can skip
printing instruction counts. You can use this verbose option as follows
- With verbose level 0, prints only summary of tflite
- With verbose level 1(default), prints all like current default except for
instructions
- With verbose level 2, prints all like current default including for
instructions

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
5 years ago[neurun] Introduce FirstfitAllocator (#3458)
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 6 Nov 2018 09:15:59 +0000 (18:15 +0900)]
[neurun] Introduce FirstfitAllocator (#3458)

* [neurun] Introduce FirstfitAllocator

Introduces FirstfitAllocator instead of BumpAllocator. This allocator
allocates a block for tensor at the first-fit offset.

Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
* Revise allocate() of FirstfitAllocator

- Adds comment for description of allocate()
- Removes inner lambda allocateBlock() in allocate()

* Change the name from Firstfit to FirstFit

* Fix typo

* Adds comment for _alloc_table

5 years agoIntroduce a enum class declaration for reduce operations. (#3461)
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 6 Nov 2018 08:50:30 +0000 (17:50 +0900)]
Introduce a enum class declaration for reduce operations. (#3461)

This commit introduces a enum declaration for reduce operations.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
5 years agoIntroduce registerSubTensorInfo (#3474)
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 6 Nov 2018 02:48:23 +0000 (11:48 +0900)]
Introduce registerSubTensorInfo (#3474)

Introduce registerTensorInfo to register subtensor information (not used yet)
Backend allocator will use this info for allocation

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
5 years agoIntroducing NotEqual op in NeuralNetworksEx (#3473)
Siva Sai Vaddipati/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 6 Nov 2018 02:30:37 +0000 (08:00 +0530)]
Introducing NotEqual op in NeuralNetworksEx (#3473)

This commit introduces NotEqual op in NeuralNetworksEx.h
Related issues: #3335, #3337

Signed-off-by: Siva Sai <siva.sai@samsung.com>
5 years agoAdd CL kernel call for Transpose op from runtime (#3470)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 6 Nov 2018 01:05:23 +0000 (06:35 +0530)]
Add CL kernel call for Transpose op from runtime (#3470)

This patch adds cl kernel call for transpose op from PACL.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoUpdating comment to create more formal equation (#3466)
Shubham Gupta/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 6 Nov 2018 00:42:06 +0000 (06:12 +0530)]
Updating comment to create more formal equation (#3466)

This patch will updates comment for sqrt so as to remove x and y and to
have more formal equation.

Signed-off-by: shubham <shub98.gupta@samsung.com>
5 years agoEnable broadcast support for Equal_Ex op (#3431)
Prasanna R/System SW /SRI-Bangalore/Engineer/삼성전자 [Tue, 6 Nov 2018 00:41:53 +0000 (06:11 +0530)]
Enable broadcast support for Equal_Ex op (#3431)

This patch enables broadcast support for Equal_Ex op.
Related issue: #3295.

Signed-off-by: prasannar <prasanna.r@samsung.com>