Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 30 Nov 2018 07:38:52 +0000 (13:08 +0530)]
Add Quant8 support to NotEqual CL (#3754)
This patch will provide quant8 support for NotEqual Op
Signed-off-by: shubham <shub98.gupta@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 30 Nov 2018 07:38:29 +0000 (16:38 +0900)]
Handle boolean for nnapi delegate (#3775)
To handle boolean for nnapi delegate, change type to quantized tensor and call NNAPI
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 30 Nov 2018 07:38:08 +0000 (13:08 +0530)]
Define Abs_Ex op in NeuralNetworksEx.h (#3758)
This patch will define Abs_Ex op in NeuralNetworksEx.h
Signed-off-by: shubham <shub98.gupta@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 30 Nov 2018 06:58:49 +0000 (15:58 +0900)]
Move benchmark into contrib (#3774)
Move benchmark into contrib
Revise directory structure: use src directory
Revise cmake to reference src directory and check build option
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Prasanna R/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 30 Nov 2018 06:23:07 +0000 (11:53 +0530)]
Introduce LOGICAL_AND op in runtime (#3564)
This patch introduces LOGICAL_AND op in our runtime.
Related issue: #3459
Signed-off-by: prasannar <prasanna.r@samsung.com>
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 30 Nov 2018 06:08:11 +0000 (15:08 +0900)]
Add delegate for SQRT and RSQRT (#3742)
* Add delegate for SQRT and RSQRT
With tensorflow v1.12, SQRT and RSQRT are built-in operator.
This commit adds delegate for built-in SQRT and RSQRT.
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
* Add macro to avoid GBS build fail
tizen(obs) has tensorflow v1.9 while this commit requires v1.12.
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 30 Nov 2018 04:13:08 +0000 (13:13 +0900)]
Introduce optional operand set on frontent (#3751)
Introduce unordered_set field for optional operand
Define API to insert optional operand
Introduce process to handle optional operand at model finish (cannot handle yet)
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 30 Nov 2018 04:10:00 +0000 (13:10 +0900)]
[neurun] Replace ITensor to our own (#3739)
Related : #3273
Part of : #3700
This commit replaces `arm_compute::ITensor` to `neurun::operand::ITensor` finally.
Replaced below classes
- ITensor
- ICLTensor
- CLTensor
- CLSubTensor
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
Prasanna R/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 30 Nov 2018 02:17:27 +0000 (07:47 +0530)]
Implement CL Kernel for ReduceMin op (#3744)
This patch implements CL Kernel for ReduceMin op.
Signed-off-by: prasannar <prasanna.r@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 30 Nov 2018 02:07:32 +0000 (07:37 +0530)]
Correction in SpaceToBatch CPU version for quant8 datatype (#3686)
This patch will set correct zero_value for quant8 datatype in spaceToBatch.
Signed-off-by: shubham <shub98.gupta@samsung.com>
Devansh Bansal/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 30 Nov 2018 01:08:48 +0000 (06:38 +0530)]
Adding nnapi delegation changes regarding TransposeConv (#3427)
This patch adds required changes to nnapi delegate to run transpose conv
Signed-off-by: b.devansh <b.devansh@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Thu, 29 Nov 2018 22:19:49 +0000 (03:49 +0530)]
Remove passing pad_quant8 gtest from skiplist (#3694)
This patch will remove pad_quant8 gtest from skiplist
Signed-off-by: shubham <shub98.gupta@samsung.com>
Prasanna R/SNAP /SRI-Bangalore/Engineer/삼성전자 [Thu, 29 Nov 2018 22:19:35 +0000 (03:49 +0530)]
Add GTest for ReduceMin op. (#3732)
This patch adds GTest for reduce_min op.
Added files:
```
reduce_min_ex
reduce_min_ex_float_1
reduce_min_ex_float_2
```
Signed-off-by: prasannar <prasanna.r@samsung.com>
Prasanna R/SNAP /SRI-Bangalore/Engineer/삼성전자 [Thu, 29 Nov 2018 06:38:18 +0000 (12:08 +0530)]
Add GTest for LogicalAND op (#3567)
This patch adds GTest for LogicalAND op.
Related issue: 3459
Signed-off-by: prasannar <prasanna.r@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Thu, 29 Nov 2018 00:58:12 +0000 (06:28 +0530)]
Add validation function & missing broadcast related code in NOTEQUAL (#3715)
This patch will add validation function and missing broadcast related code.
Signed-off-by: shubham <shub98.gupta@samsung.com>
Prasanna R/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 28 Nov 2018 10:43:47 +0000 (16:13 +0530)]
Introduce ReduceMin in PACL (#3729)
This patch introduces ReduceMin in PACL.
Related issue: #3459
Signed-off-by: prasannar <prasanna.r@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 28 Nov 2018 09:34:23 +0000 (18:34 +0900)]
[neurun] Introduce Node.Include.h (#3740)
This file includes headers of all the nodes
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 28 Nov 2018 09:18:25 +0000 (18:18 +0900)]
[neurun] Change dot_dumper input/output shape (#3741)
This commit changes `input`/`output` shape of dot_dumper graph to `doublecircle`.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 28 Nov 2018 07:58:19 +0000 (13:28 +0530)]
Removing the not required Pad op files (#3625)
After PR #3428 merged
These PadLayer files are not required.
Signed-off-by: shubham <shub98.gupta@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 28 Nov 2018 06:42:17 +0000 (15:42 +0900)]
Update build-docker command (#3697)
- Change command: build_docker -> build-docker
- Can use docker build options
- Update docker file: restrict yapf version to 0.22.0
- Update document
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 28 Nov 2018 04:54:13 +0000 (13:54 +0900)]
Save pointer to buffer for operand value setting (#3736)
For operand value setting, save pointer to buffer instead of copy when size is greater than
ANEURALNETWORKS_MAX_SIZE_OF_IMMEDIATELY_COPIED_VALUES defined in NeuralNetworks.h
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 28 Nov 2018 04:22:14 +0000 (09:52 +0530)]
Correction in pad CPU version for quant8 datatype (#3687)
This patch set the correct padding value which is offset for quant8 dataypes.
Signed-off-by: shubham <shub98.gupta@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 28 Nov 2018 01:34:54 +0000 (07:04 +0530)]
Add CLkernel for Generic Padding (#3428)
This patch will add cl kernel to execute pad op on GPU
Signed-off-by: shubham <shub98.gupta@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 28 Nov 2018 01:33:45 +0000 (10:33 +0900)]
Support Layout::NHWC for CL Kernel of SpaceToBatchND (#3734)
This commit supports Layout::NHWC for CL Kernel of SpaceToBatchND.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 28 Nov 2018 01:04:11 +0000 (10:04 +0900)]
Apply CL Kernel of SpaceToBatchND to PACL (#3731)
This commit applies CL Kernel of SpaceToBatchND to PACL.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 28 Nov 2018 01:03:46 +0000 (06:33 +0530)]
Remove not req headers from NotEqual cl files (#3728)
This patch will remove not required headers included
in src/runtime/CL/functions/CLNotEqual.cpp
Signed-off-by: shubham <shub98.gupta@samsung.com>
Vishal Keshav/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 28 Nov 2018 00:57:32 +0000 (06:27 +0530)]
Added implementation for Pack operation with ACL (#3398)
Implements 3D to 4D tensor pack operation
Uses ACL kernels for packing tensors
Signed-off-by: Vishal Keshav <vishal.k1@samsung.com>
Prasanna R/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 28 Nov 2018 00:54:24 +0000 (06:24 +0530)]
Update Validation of Arguments in SquaredDifference CL kernel (#3433)
This patch will update validation of arguments in CL Kernel of SquaredDifference op.
Signed-off-by: prasannar <prasanna.r@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 28 Nov 2018 00:53:41 +0000 (09:53 +0900)]
Workaround to pass scalar type tensor in nnapi delegate (#3707)
Some model have scalar type input/output, but NNAPI requre tensor type for these operands. (ex. elementwise sub, div)
To satisfy NNAPI spec, change scalar to vector with size 1 in nnapi delegate.
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Prasanna R/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 28 Nov 2018 00:44:46 +0000 (06:14 +0530)]
Introduce ReduceMin in the enum class declared for reduce ops (#3733)
This patch introduces ReduceMin in the enum class declared for reduce ops.
Signed-off-by: prasannar <prasanna.r@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 27 Nov 2018 22:11:51 +0000 (07:11 +0900)]
[neurun] Introduce CLTensor/CLSubTensor (#3723)
Related : #3273
Part of : #3700
This commit introduces `neurun::backend::acl_cl::operand::CLTensor`/`neurun::backend::acl_cl::operand::CLSubTensor`.
Those are inherited to internal `ICLTensor`, and have handles of `ACL` `arm_compute::CLTensor` and `arm_compute::CLSubTensor`.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 27 Nov 2018 22:11:41 +0000 (07:11 +0900)]
[neurun] Fix Graph Verifier visit check (#3722)
Visit and OnStack checking assumed ID range as vector which is wrong.
This commit changes variable `visit` and `on_stack` to be a map so it
can correctly check nodes for sure.
In the same manner with #3721
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
Prasanna R/SNAP /SRI-Bangalore/Engineer/삼성전자 [Tue, 27 Nov 2018 10:34:57 +0000 (16:04 +0530)]
Introduce REDUCE_MIN_EX in NeuralNetworksEx.h (#3570)
This patch introduces REDUCE_MIN_EX in NeuralNetworksEx.h.
Related issue: #3459
Signed-off-by: prasannar <prasanna.r@samsung.com>
[neurun] Remove permutation layer for model's input/output and permute in-place (#3617)
Permute while copying from/to interpreter: this will eliminate extra
copying in permutation layer. This is just for NHWC_TO_NCHW in input
and NCHW_TO_NHWC in output.
Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 27 Nov 2018 10:33:46 +0000 (19:33 +0900)]
Implement CL kernel of SpaceToBatchND for NCHW (#3633)
This commit implements CL kernel of SpaceToBatchND for NCHW.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Tue, 27 Nov 2018 10:31:42 +0000 (16:01 +0530)]
Rearranging rsqrt gtest according to the name in all_generated_tests.cpp (#3716)
When generate_test script is executed this rsqrt declaration will get moved
to the different line in all_generated_tests file based on the sorting done
with name.
This patch will update the all_generated_tests.cpp in master so as to
have same code as generated by generate_test script
Signed-off-by: shubham <shub98.gupta@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 27 Nov 2018 08:43:49 +0000 (17:43 +0900)]
[neurun] Introduce ITensor/ICLTensor (#3719)
* [neurun] Introduce ITensor/ICLTensor
This commit introduces `neurun::backend::operand::ITensor`/`neurun::backend::acl_cl::operand::ICLTensor` since we want to define internal interface of `ITensor` that is an common interface of all backends.
In case of `ACL` backend, It should be defined to `ACL` wrapper classes inherited to internal `ICLTensor`.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
* Include some functions of derived classes
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 27 Nov 2018 07:32:12 +0000 (16:32 +0900)]
Add generated tests of TensorflowSum for 4-D (#3388)
This commit adds generated tests of TensorflowSum(ReduceSum) for 4-D.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 27 Nov 2018 07:31:55 +0000 (16:31 +0900)]
Add 4 tflite models for framework test (#3720)
* Add 4 tflite models for framework test
exp_4d, neg_4d, rsqrt_4d, sqrt_4d tflite models are added into
framework test.
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
* Disable sqrt_4d and rsqrt_4d model test
since delegates for SQRT and RSQRT are not prepared
* Remove comments
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Tue, 27 Nov 2018 06:46:29 +0000 (12:16 +0530)]
Correction of selected op in notequal gtest (#3714)
This patch will set notequal op name instead of equal op name
in notequal gtest
Signed-off-by: shubham <shub98.gupta@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Tue, 27 Nov 2018 06:28:46 +0000 (15:28 +0900)]
[neurun] Fix Dfs Iterator (#3721)
As mentioned from #3617 by @d-poshshoev, Visit checking assumed ID range
as vector which is wrong. This commit changes variable `visit` to be a
map so it can correctly check nodes for sure.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Tue, 27 Nov 2018 04:13:42 +0000 (13:13 +0900)]
Fix incorrect path of lcov_cobertura.py in script (#3718)
There was a typo on sciprt path for lcov_cobertura.py
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Tue, 27 Nov 2018 01:23:50 +0000 (06:53 +0530)]
Add missing unpak gtest declaration from all_generated_test and skiplist (#3691)
This patch will add missing unpack gtest declaration from all_generated_test.cpp and skiplist
Signed-off-by: shubham <shub98.gupta@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Tue, 27 Nov 2018 01:13:42 +0000 (10:13 +0900)]
Don't run MODELS/* in run_benchmark_op.sh (#3709)
`run_benchmark_op.sh` is for benchmarking a given operator.
Models such as inception_nonslim, inception_slim and mobilenet are
located under MODELS, and executed from `run_benchmark_op.sh`.
Thus, `run_benchmark_op.sh` excludes models under `MODELS`.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
Devansh Bansal/SNAP /SRI-Bangalore/Engineer/삼성전자 [Tue, 27 Nov 2018 01:01:25 +0000 (06:31 +0530)]
Generate NNAPI Test Case(a.k.a generatedtests) for Split (#3683)
This PR adds generatedtests for Split. #3459
Signed-off-by: b.devansh <b.devansh@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Mon, 26 Nov 2018 11:35:53 +0000 (17:05 +0530)]
Add spaceToBatchND gtests to test op for quant8 dataype (#3688)
This patch adds gtest to test SpaceToBatchND op for quant8 dataype
Signed-off-by: shubham <shub98.gupta@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 26 Nov 2018 11:20:40 +0000 (20:20 +0900)]
Fix tflite_benchmark_model out-of-memory error (#3706)
Previously, benchmark_model created nnapi_delegate every run.
Now, nnapi_delegate build graph only once.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 11:13:20 +0000 (20:13 +0900)]
Remove unused CL Kernels of Reduce Operation (#3702)
This commit removes CL Kernels of Reduce Operation that is no longer needed.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 10:55:40 +0000 (19:55 +0900)]
Apply CL Kernel of HashtableLookup for PACL (#3701)
This commit applies CL Kernel of HashtableLookup for PACL.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 08:13:44 +0000 (17:13 +0900)]
Add generated tests of Mean for 4-D (#3453)
This commit Add generated tests of Mean for 4-D.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
윤지영/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 07:47:31 +0000 (16:47 +0900)]
Do not use info directory in tflkit (#3647)
Move the *.template files in tflkit into its parent dir
And ignore .info files in tflkit
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 26 Nov 2018 06:08:36 +0000 (15:08 +0900)]
Move lcov-to-cobertura-xml script into externals (#3693)
Move lcov-to-cobertura-xml script into externals
Recover script file name to original: lcov_cobertura.py
Change imported_url.txt to README.md
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 26 Nov 2018 05:57:04 +0000 (14:57 +0900)]
Update include order and comment in Compiler.cc (#3684)
Update include order: categorize as
- header for Compiler.cc
- header in same directory
- header in differnt neurun directory
- header in include
Add and update comment
- Update comment for SubTensorAnlyzer
- Add comment for compilation phase (backend independent, lower, backend dependent, codegen)
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 26 Nov 2018 05:56:46 +0000 (14:56 +0900)]
Fix wrong variable name in test_driver.sh (#3699)
I've found report_dir does not pass properly during investigation of #3420.
Then, I found a typo of a vaiable name in test_driver.sh script.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
이상규/동작제어Lab(SR)/Principal Engineer/삼성전자 [Mon, 26 Nov 2018 05:56:12 +0000 (14:56 +0900)]
tflite_benchmark_model is updated to v1.12.0. (#3660)
Most files are not changed from v1.12.0.
My modification is to support multiple kernel expansion operators.
You can find the changes from stats_calculator.cc and profile_summarizer.cc.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 05:43:58 +0000 (14:43 +0900)]
Apply CL Kernel of ReduceOperation to PACL. (#3673)
This commit applies CL Kernel of ReduceOperation to PACL.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 05:40:36 +0000 (14:40 +0900)]
Implement CL kernel of HashtableLookup (#3500)
This commit implements CL kernel of HashtableLookup op.
- Add an IFunction class for HashtableLookup.
- Add an ICLKernel class for HashtableLookup.
- Add cl kernel functions for HashtableLookup.
- Append the cl functions to kernel_program_map.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 02:38:29 +0000 (11:38 +0900)]
[neurun] Remove ptr_to_element in ConstantInitializer (#3696)
Related : #3273
This commit remove using `ptr_to_element` in `ConstantInitializer` for removing `arm_compute` dependency.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 01:19:28 +0000 (10:19 +0900)]
[model_parser] Print memory size of each tensor (#3677)
Now model_parser prints memory size of each tensor. In addition,
expected allocating memory also prints.
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Mon, 26 Nov 2018 01:19:01 +0000 (10:19 +0900)]
[neurun] Simplify SubTensorAnalyzer (#3690)
Make SubTensorAnalyzer to Override only ConcatNode.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 26 Nov 2018 01:12:25 +0000 (10:12 +0900)]
Make script/config directory (#3692)
Make script/config directory and move configuration files for script into script/config
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Mon, 26 Nov 2018 01:11:57 +0000 (06:41 +0530)]
Correction in Padding_Same untility function (#3629)
This patch will correct the Paddin_Same function logic.
Signed-off-by: shubham <shub98.gupta@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Mon, 26 Nov 2018 01:06:56 +0000 (10:06 +0900)]
Delete docker_helper file (#3689)
Delete docker_helper script file: not used any more
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
[neurun] Add PermutationEliminationPass to eliminate permutation of I/O (#3552)
Eliminates permutation after model's input and before output
This is just for NHWC_TO_NCHW in input and NCHW_TO_NHWC in output.
Permutation of input/output will be done during source/sink in next commit
Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 23 Nov 2018 06:05:32 +0000 (15:05 +0900)]
[neurun] Auto-generate NodeVisitor headers (#3685)
Auto-generate NodeVisitor headers with macro.
Resolve #3664
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
김용섭/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 23 Nov 2018 06:01:12 +0000 (15:01 +0900)]
[neurun] Doxygen comments for MemoryPlanner on CPU (#3679)
* [neurun] Doxygen comments for MemoryPlanner on CPU
Doxygen comments for MemoryPlanner on CPU
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
* Remove @ingroup
* Fix typo. opernad -> operand
Vishal Keshav/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 23 Nov 2018 02:38:54 +0000 (08:08 +0530)]
Fixed the test cases for CUSTOM pack operations (#3640)
Input tensors are treated as buffer, and packing
will be done with strides determined by axis.
Signed-off-by: Vishal Keshav <vishal.k1@samsung.com>
Vishal Keshav/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 23 Nov 2018 02:38:40 +0000 (08:08 +0530)]
Defined CUSTOM pack operation (#3648)
Added details for pack operation
Signed-off-by: Vishal Keshav <vishal.k1@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Fri, 23 Nov 2018 02:38:02 +0000 (08:08 +0530)]
Correction in pad_quant gtest (#3680)
This patch will correct the output buffer values and name of the pad_quant gtest.
Signed-off-by: shubham <shub98.gupta@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Fri, 23 Nov 2018 01:38:33 +0000 (10:38 +0900)]
Revise coverage build and report script (#3661)
Revise docker_cross_test_coverage_build.sh and docker_coverage_report.sh
- Remove using nnfw_docker file
- Remove deprecated environment variable setting
- Use docker-run-user command
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Fri, 23 Nov 2018 01:21:10 +0000 (10:21 +0900)]
[neurun] Extend Op.lst macro (#3681)
Make `Op.lst` always keep the full list of operations. Add a boolean
parameter to determine if it is from NN API or internally used node.
Later, this list will be used for auto generation of visitor code.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
[neurun] Add remove() for operand::Set and operation::Set and Graph (#3678)
These changes are needed to remove some layers from a graph
Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 22 Nov 2018 09:25:38 +0000 (18:25 +0900)]
[neurun] Introduce internal TensorInfo class (#3674)
Related : #3273
Part of : #3484
This commit introduces internal `TensorInfo` class that has just `Shape` and `TypeInfo`.
Entire dependency of `arm_compute::TensorInfo` for `cpu` backend will be reduced after cpu `Tensor` class revised in another PR.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 22 Nov 2018 09:25:26 +0000 (18:25 +0900)]
Move Plan and related codes into compiler (#3676)
Move Plan and related codes into compiler
Update namespace and comments
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
김수진/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 22 Nov 2018 08:21:34 +0000 (17:21 +0900)]
[neurun] Introduce internal IFunction class (#3672)
* [neurun] Introduce internal IFunction class
Related : #3273
Part of : #3484
This commit introduces internal `IFunction` class.
Signed-off-by: sjsujinkim <sjsujin.kim@samsung.com>
* Change directory
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 22 Nov 2018 08:20:35 +0000 (17:20 +0900)]
Support generate rootfs for target os ubuntu 18.04 (#3675)
Support generate rootfs for target os ubuntu 18.04 (ex. xu4 ubuntu mate 18.04)
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Vishal Keshav/SNAP /SRI-Bangalore/Engineer/삼성전자 [Thu, 22 Nov 2018 04:59:12 +0000 (10:29 +0530)]
Fixed unpack node inputs (#3639)
The inputs buffer contains number of splits
Following tf_ops_compatibility
Signed-off-by: Vishal Keshav <vishal.k1@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 22 Nov 2018 04:45:19 +0000 (13:45 +0900)]
Move BackendResolver into compiler (#3671)
Move BackendResolver into compiler and fix namespace
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 22 Nov 2018 04:41:08 +0000 (13:41 +0900)]
[neurun] Rename operation Node files (#3668)
Rename operation Nodes from `{NodeName}.h` to `{NodeName}Node.h` so
they can match with their class names.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 22 Nov 2018 03:49:23 +0000 (12:49 +0900)]
Move ConstantInitializer into compiler (#3670)
Move ConstantInitializer into compiler and fix namespace
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Thu, 22 Nov 2018 03:07:51 +0000 (08:37 +0530)]
Introduce an enum class declaration for logical operations (#3624)
To have unified cl kernels for logical operations,
This commit introduces a enum declaration for logical operations.
Signed-off-by: shubham <shub98.gupta@samsung.com>
이춘석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Thu, 22 Nov 2018 01:45:57 +0000 (10:45 +0900)]
Update nnfw's register.cc to v1.12.0 (#3607)
- Add v1.12 missing operators into register.cc
- FIXME: In obs build, do not apply this patch
Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 22 Nov 2018 01:45:34 +0000 (10:45 +0900)]
[neurun] Remove NOP node (#3665)
NOP node was first introduced for general testing purpose but it does
not used or necessary any longer.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
[neurun] Define PermutationEliminationPass class (#3652)
This class eliminates permutation of input/output
just for NHWC_TO_NCHW in input and NCHW_TO_NHWC in output.
Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
장지섭/동작제어Lab(SR)/Engineer/삼성전자 [Thu, 22 Nov 2018 00:56:44 +0000 (09:56 +0900)]
Implement unified CL Kernel of Reduce operations into ARMComputeEx (#3490)
This commit implements unified CL Kernel of Reduce operations into ARMComputeEx.
- Introduce class CLReduceOperation.
- Introduce class CLReduceOperationKernel.
- Add cl functions for Reduce operations to reduce_operation.cl.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Thu, 22 Nov 2018 00:54:16 +0000 (06:24 +0530)]
Remove transpose gtests from skiplist (#3471)
This patch will remove transpose gtests from skiplist
Signed-off-by: shubham <shub98.gupta@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 21 Nov 2018 10:29:56 +0000 (19:29 +0900)]
Move PlanBuilder into compiler (#3667)
Move PlanBuilder.cc/h into compiler and fix namespace
Remove IPlanBuilder interface
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 21 Nov 2018 09:32:25 +0000 (18:32 +0900)]
Move SubTensorInfo into compiler (#3666)
Move SubTensorInfo.h into compiler and fix namespace
Remove acl dependency header from SubTensorInfo.h
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Vishal Keshav/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 21 Nov 2018 09:15:26 +0000 (14:45 +0530)]
Added defenition for CUSTOM unpack ops (#3650)
Defined the constraints for unpack ops
Signed-off-by: Vishal Keshav <vishal.k1@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 21 Nov 2018 08:27:17 +0000 (17:27 +0900)]
Move OperationValidator into compiler (#3662)
Move OperationValidator into compiler directory
Fix namespace
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이한종/동작제어Lab(SR)/Engineer/삼성전자 [Wed, 21 Nov 2018 08:00:55 +0000 (17:00 +0900)]
[neurun] Graph holds Model as pointer (#3663)
Graph holds Model as pointer. This is for the next work that `_model` is moved
to Linear.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 [Wed, 21 Nov 2018 06:29:26 +0000 (15:29 +0900)]
Merge middleend into compiler directory (#3658)
Move SubTensorAnalyzer.cc/h into compiler directory
Fix namespace and cmake
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Shubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 [Wed, 21 Nov 2018 06:27:35 +0000 (11:57 +0530)]
Removing unused variables and rearranging by name in TransposeConv (#3656)
This patch will remove unused variables and rearranges the variables by their name
Signed-off-by: shubham <shub98.gupta@samsung.com>
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>
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>
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>
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>
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>
NeuralNetworkEx code cleanup (#3649)
Just little code cleanup for NeuralNetworkEx.h
Signed-off-by: Tanuj Tekriwal <tanuj.tekri@samsung.com>
[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>