이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 3 Sep 2019 08:23:17 +0000 (17:23 +0900)]
[tflite_loader] implement squared_difference (#7127)
It adds squared_difference operator in tflite_loader.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 3 Sep 2019 08:12:48 +0000 (17:12 +0900)]
Enable L2Normalization op for ACL neon (#7129)
This commit enables to support L2Normalization op for ACL neon.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 3 Sep 2019 08:12:14 +0000 (17:12 +0900)]
Fix misapplied enum value in KernelGenerators (#7130)
This commit fixes misapplied enum value in KernelGenerators.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 3 Sep 2019 08:11:22 +0000 (17:11 +0900)]
Enable acl neon backend mul operation (#7105)
Update acl neon backend ShapeFixer's mul
Remove exception in kernel generator for mul
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 3 Sep 2019 06:52:46 +0000 (15:52 +0900)]
[tflite_loader] implement tanh (#7126)
It adds tanh operator in tflite_loader.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 3 Sep 2019 05:50:11 +0000 (14:50 +0900)]
Remove hardcode for L2Normalization (#7122)
This commit removes hardcode for L2Normalization from KernelGenerator of acl_cl
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 3 Sep 2019 05:49:01 +0000 (14:49 +0900)]
[tflite_loader] implement rsqrt (#7124)
It adds rsqrt operator in tflite_loader.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 3 Sep 2019 05:48:07 +0000 (14:48 +0900)]
Change to include the integrated header of ACL NEON (#7117)
This commit changes to include the integrated header of ACL NEON.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
윤지영/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 3 Sep 2019 05:46:42 +0000 (14:46 +0900)]
[neurun] Add srcn library for new neurun backend (#6993)
* Add srcn library for new neurun backend
This patch includes implementation of srcn kernel.
It only supports Convolution and TransposeConv operations.
Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
* Fix format checker errors
* Remove unused code and unnecessary comments
* Remove ${} from REMOVE_ITEM in cmakelist file
* Do not build srcn library on x86_64 machine
* Add return case for OpenMP library not installed
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 3 Sep 2019 05:39:38 +0000 (14:39 +0900)]
[neurun] Appends apis to ITensorBuilder (#6973)
* [neurun] Appends apis to ITensorBuilder
Appends apis to ITensorBuilder for deallocating constant tensors if they
are unused and separating constant tensors and nonconstant tensors
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
* Change afterFunctionPrepare to postFunctionPrepare
* Fix default parameter of registerTensorInfo
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 3 Sep 2019 04:52:29 +0000 (13:52 +0900)]
[neurun] Make {pre|post}Visit of TemplTensorBuilder recursive (#7064)
Make {pre|post}Visit of TemplTensorBuilder recursive to support for more
than two consecutive concat layers
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 3 Sep 2019 04:33:20 +0000 (13:33 +0900)]
Update dockerfile - numpy (#7111)
Install python3 numpy for andorid-q test generator
Change python-pip to python3-pip
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 3 Sep 2019 02:24:21 +0000 (11:24 +0900)]
Use static casting to convert enum type (#7104)
To resolve implicit enum type casting warning, introduce convert helper and use static casting
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Tue, 3 Sep 2019 01:08:24 +0000 (10:08 +0900)]
[res] Introduce UNIT_Mean_00X testcases (#7051)
This commit will introduce `UNIT_Mean_000` ~ `UNIT_MEAN_004`
testcases
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Tue, 3 Sep 2019 01:03:04 +0000 (10:03 +0900)]
[exo-tflite] Adding TFLShapeInferenceRule into ShapeInference (#7096)
* [exo-tflite] Adding TFLShapeInferenceRule into ShapeInference
This adds TFLShapeInferenceRule into ShapeInference.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* restore empty line
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 3 Sep 2019 00:41:42 +0000 (09:41 +0900)]
Explicit type to prevent overflow (#7106)
To remove integer overflow warning by svace, use explicit type int64_t, not auto
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 3 Sep 2019 00:38:54 +0000 (09:38 +0900)]
Fix copy constructor (#7102)
Fix uninitialized field in copy constructor
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Tue, 3 Sep 2019 00:36:24 +0000 (09:36 +0900)]
Fix deref null pointer error on scheduler (#7094)
Change to runtime error instead of assert for deref null pointer error on scheduler
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 22:30:59 +0000 (07:30 +0900)]
[exo-tflite] Introducing TFLMaxPool2D (#7087)
This adds TFLMaxPool2D into TFLDialect.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 2 Sep 2019 22:20:14 +0000 (07:20 +0900)]
[moco-tf] Bugfix of FixShapeTransform of Squeeze (#7031)
If `Squeeze` operation have `squeeze_dims` as `[-1,-2]` and input rank is 4, it should be converted to `[2, 3]`
According to current implementation, the code erase -1 first and then put 3 in the `squeeze_dims`
However, it causes some bug. If -1 is erased, iterator of `squeeze_dims` points -2.
And then if 3 is inserted, the iterator points 3 because it is `Set`!
Now, `squeeze_dims` is `[-2, 3]` and iterator try to iterate next element but there is no more element
because 3 is the last element.
Therefore this commit will fix this bug by introducing new `Set`
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 22:06:44 +0000 (07:06 +0900)]
[moco-tf] Update FixShapeTransform test with AvgPool (#7041)
This will update FixShapeTransform test to test with TFAvgPool
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
[mir] Remove support for Reduce and Pool (#7074)
- Reduce and Pool operations were replaced by `ReduceMean`, `AvgPool2D` and `MaxPool2D`.
- Make `ReduceOp` a base class for `ReduceMeanOp`.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir2loco] Remove support for PoolOp (#7071)
This operation was replaced by `AvgPool2DOp` and `MaxPool2DOp`.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[nnc] Remove support for Reduce and Pool (#7057)
- These operation were replaced by `ReduceMean`, `AvgPool2D` and `MaxPool2D`.
- Pool operations are supported in DataFormatSwitcher optimization.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 09:06:39 +0000 (18:06 +0900)]
[exo-tflite] functions to build node (#7080)
This adds functions to build featureEncode and featureDecode, which are frequently used.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 09:04:59 +0000 (18:04 +0900)]
[exo-tflite] Adding knob for AveragePool2D conversion (#7092)
This adds knob for AveragePool2D conversion. Default is false.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 2 Sep 2019 08:25:52 +0000 (17:25 +0900)]
Enable Floor op for ACL neon (#7100)
This commit enables to support Floor op for ACL neon.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 08:24:26 +0000 (17:24 +0900)]
[exo-tflite] Adding TFLTypeInferenceRule into TypeInference (#7097)
This adds TFLTypeInferenceRule into TypeInference.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 07:55:58 +0000 (16:55 +0900)]
[tflite_loader] Add Mul and Sub operator (#7099)
It adds mul and sub operator to run ST model.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 2 Sep 2019 06:54:24 +0000 (15:54 +0900)]
Fix uninitialized field (#7089)
Fix uninitialized field contructor warning in neurun core and backend
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 2 Sep 2019 06:14:04 +0000 (15:14 +0900)]
Enable Exp op for ACL neon (#7093)
This commit enables to support Exp op for ACL neon.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 04:16:38 +0000 (13:16 +0900)]
[exo-tflite] Change the name of input of TFLRelu (#7081)
* [exo-tflite] Change the name of input of TFLRelu
Input name follows TensorFlow input naming. (input -> features)
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* modify input -> features
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Mon, 2 Sep 2019 04:08:32 +0000 (13:08 +0900)]
Apply finding frontend layout of operand (#7065)
This commit applies finding frontend layout of operand.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 04:08:03 +0000 (13:08 +0900)]
[tflite_loader] implement Relu6 operator (#7068)
It adds `Relu6` for mobilenet v3.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 2 Sep 2019 03:45:24 +0000 (12:45 +0900)]
Fix null type reference (#7059)
Instead of using getOperandSize, use type pointer to find out operand has unspecified shape
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Mon, 2 Sep 2019 03:44:12 +0000 (12:44 +0900)]
[exo-tflite] Introducing TFLAveragePool2D node (#7054)
* [exo-tflite] Introducing TFLAveragePool2D node
This adds TFLAveragePool2D node, which represelt AVERAGE_POOL_2D of TensorFlow Lite.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* modified per jh1304's review. Filter and Stride were made
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Mon, 2 Sep 2019 00:56:33 +0000 (09:56 +0900)]
Coverity issue on scheduler (#7072)
- Unreached default in switch-case
- Unchecked return of library call
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 1 Sep 2019 23:03:02 +0000 (08:03 +0900)]
[exo-tflite] Introducing TFLTypeInference (#7015)
This adds TFLTypeInference to infer type for TFL nodes.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Sun, 1 Sep 2019 23:02:26 +0000 (08:02 +0900)]
[pepper-str] Fix "unused parameter" warning in CI (#7075)
This fix "unused parameter" warning in CI.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 30 Aug 2019 09:59:01 +0000 (18:59 +0900)]
Remove unnecessary helper function (#7066)
This commit removes unnecessary helper function.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 30 Aug 2019 09:41:08 +0000 (18:41 +0900)]
Make ConcatLayers support NCHW frontend layout (#7063)
This commit makes ConcatLayers support NCHW frontend layout.
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
이한종/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 30 Aug 2019 09:25:38 +0000 (18:25 +0900)]
[neurun] Use LIB_PTHREAD variable in CMake (#7069)
Use variable `${LIB_PTHREAD}` rather than hardcoded `pthread`. This is
useful for Android build which does not have pthread lib. Android
contains pthread in libc.
Signed-off-by: Hanjoung Lee <hanjoung.lee@samsung.com>
장지섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 30 Aug 2019 08:37:25 +0000 (17:37 +0900)]
Revert commits related to mutex in acl_cl::Object (#7062)
- Revert "Fix deadlock when access two acl_cl at once (#6500)"
- Revert "[neurun] Change acl_cl::Object::access to multithread safe (#5636)"
This reverts commit
51671b098692a133d86624afc5455c6e01ef9302
This reverts commit
6097ce0e7d59be4342db40e1bfc104d5ec88c5fa
Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
[mir_caffe] Switch to AvgPool2D and MaxPool2D ops (#7029)
Switch from `PoolOp` to `AvgPool2DOp` and `MaxPool2DOp`. The former ones are deprecated.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 30 Aug 2019 07:32:17 +0000 (16:32 +0900)]
Check dladdr return value (#7061)
Check dladdr return value and log if fail to load
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 30 Aug 2019 07:17:14 +0000 (16:17 +0900)]
Fix unused return value (#7060)
Fix unused return value of call for dimension valid check
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
[mir_onnx] Switch GlobalAveragePool converter to AvgPool2DOp (#7049)
Switch from `PoolOp` to `AvgPool2DOp` in `GlobalAveragePool` converter. The former one is deprecated.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 30 Aug 2019 07:07:32 +0000 (16:07 +0900)]
[exo-tflite] Make "Check.h" to use pepper instead of pp (#7038)
* [exo-tflite] added pp into cmake scripts
`pp` is added into cmake scripts to use "Check.h", which includes a header in pp.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* pp to lightweight pepper::str
[mir] Remove operation names (#7016)
Names on operations were difficult to maintain.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
오형석/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 30 Aug 2019 05:49:30 +0000 (14:49 +0900)]
Fix uncatched exception (#7058)
Operand::operandSize() can throw exeception because sizeOfDataType() in DataType.h can throw runtime exception
Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 30 Aug 2019 05:02:07 +0000 (14:02 +0900)]
[neurun] Resolve Division by zero in Softmax layer (#7055)
This commit will resolve division by zero in Softmax layer
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 30 Aug 2019 02:41:38 +0000 (11:41 +0900)]
[pepper-str] make a function in header inline (#7043)
* [pepper-str] make a function in header inline
This adds `inline` keyword to `str_impl(...)` function in header.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* added inline to all str_impl
Ivan Vagin/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 30 Aug 2019 02:17:48 +0000 (11:17 +0900)]
[neurun] Implemented scheduler test with branched graph (#6926)
* [neurun] Implemented scheduler test with branched graph
* Implemented scheduler test with branched graph and known execution time
* Added set time methods to ExecTime
* Refactored existing scheduler test a little bit
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
* 'set' methods from ExecutionTime replaced with helper functions
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
* remove profile data before test
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
* Review fixes
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
* Increased execution time for GPU in last test
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
* Set correct large exec time for gpu
Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
Ivan Vagin/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 30 Aug 2019 01:43:27 +0000 (10:43 +0900)]
[neurun] Prettified HE scheduler a little bit (#6982)
* [neurun] Prettified HE scheduler a little bit
* Removed backend transfer time from _run_cache
* _run_cache replaced with _preferred_operations
* Profiling information existence checks added
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
* Review fixes
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
Ivan Vagin/AI Tools Lab /SRR/Engineer/삼성전자 [Fri, 30 Aug 2019 01:39:28 +0000 (10:39 +0900)]
[neurun] Made scheduler to shuffle backends after all nodes are profiled (#7010)
[neurun] Made scheduler to shuffle backends after all nodes are already profiled
Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 30 Aug 2019 00:59:05 +0000 (09:59 +0900)]
[neurun] Introduce ITensorManager to be used as MemoryManager (#7048)
Introduce ITensorManager to be used as MemoryManager instead of existing
MemoryManager in the future.
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Fri, 30 Aug 2019 00:44:59 +0000 (09:44 +0900)]
[moco-tf] Delete FixPaddingTransform (#7042)
This will delete unused FixPaddingTransform
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Fri, 30 Aug 2019 00:04:32 +0000 (09:04 +0900)]
[tf2tflite] Enable moco logs via TF2TFLITE_Log_Frontend (#7040)
Let's use TF2TFLITE_Log_Frontend instead of MOCO_LOG.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
남궁석/On-Device Lab(SR)/Engineer/삼성전자 [Fri, 30 Aug 2019 00:04:20 +0000 (09:04 +0900)]
[res] Introduce UNIT_Squeeze_003 test (#7039)
This test is for checking that `Squeeze` operation handle multiple negative
squeeze_dims or not
Signed-off-by: Seok NamKoong <sk.namkoong@samsung.com>
[mir_tflite] Switch to AvgPool2D and MaxPool2D ops (#7024)
Switch from `PoolOp` to `AvgPool2DOp` and `MaxPool2DOp`. The former ones are deprecated.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_caffe2] Switch to AvgPool2D and MaxPool2D ops (#7032)
Switch from `PoolOp` to `AvgPool2DOp` and `MaxPool2DOp`. The former ones are deprecated.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_onnx] Switch to AvgPool2D and MaxPool2D ops (#7023)
Switch from `PoolOp` to `AvgPool2DOp` and `MaxPool2DOp`. The former ones are deprecated.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir2loco] Add support for AvgPool2D and MaxPool2D (#7022)
These operations are the future replacement of the `Pool` operation.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_tflite] Switch to ReduceMean op (#7004)
Switch from `ReduceOp` to `ReduceMeanOp`. The former one is deprecated.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_onnx] Switch to ReduceMean op (#7003)
Switch from `ReduceOp` to `ReduceMeanOp`. The former one is deprecated.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
Павел Ильютченко/AI Tools Lab /SRR/Engineer/삼성전자 [Thu, 29 Aug 2019 19:02:44 +0000 (22:02 +0300)]
[nnc] Implemented Data Format Switcher (#6932)
* Implemented PoolOp visitor in switcher
Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 29 Aug 2019 10:00:48 +0000 (19:00 +0900)]
[neurun] Introduce AclInternalBufferManager class (#6983)
Introduce AclInternalBufferManager class which will be used by acl's
layers
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 09:48:23 +0000 (18:48 +0900)]
[neurun] Fix nnfw_output_tensorinfo (#7035)
nnfw_session::output_tensoninfo should uses output size, instead of input size.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 09:35:02 +0000 (18:35 +0900)]
[moco-tf] Use logo Pass (#7030)
* [moco-tf] Use logo Pass
This will update to use logo Pass as transformations for two Optimiziers
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use moco::tf::Knob
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 09:20:58 +0000 (18:20 +0900)]
[moco-tf] Use loco::NodeShape for inputs (#7008)
* [moco-tf] Use loco::NodeShape for inputs
This will change fix_shape() to use loco::NodeShape for input's shape information
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* fix vars
[nnc] Add support for ReduceMean (#7002)
This operation is the future replacement of the `Reduce` operation.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[nnc] Add support for AvgPool2D and MaxPool2D (#7011)
These operations are the future replacement of the `Pool` operation.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 08:27:01 +0000 (17:27 +0900)]
[exo-tflite] Placeholders for TFL nodes (#7021)
* [exo-tflite] Placeholders for TFL nodes
This will add TODO comment as placeholders for TFL nodes to come
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use name from schema
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 08:19:11 +0000 (17:19 +0900)]
[exo-tflite] Introducing TFLShapeInference (#6991)
* [exo-tflite] Introducing TFLShapeInference
This adds TFLShapeInference that infer the shape of TFL nodes and TFLShapeAnnot that passes shape from canonical node.
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* adding missing graph input
* fix typo
* remove Forward word from internal class naming
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 29 Aug 2019 07:25:31 +0000 (16:25 +0900)]
[neurun] Add iterate() to FunctionSequence (#6977)
Add iterate() to FunctionSequence for #6156
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 07:05:30 +0000 (16:05 +0900)]
[nnpkg_tool] Revise nnpkg_test and model2nnpkg for readable log (#7028)
`nnpkg_test`
- distinguish the run and compare result.
- shows the errors after a blank line.
- does not use `set -e` to continue after error for pretty printing.
`model2nnpkg`
- It becmome less noisy.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 07:02:38 +0000 (16:02 +0900)]
[nnpkg_tool] Introduce nncc-tc-to-nnpkg-tc (#7005)
Introduce tool for converting nncc testcase to nnpkg testcase.
Please read README.md for usage.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
[mir] Add AvgPool2D and MaxPool2D operations (#6996)
These are the future replacement of the `Pool` operation.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir] Add ReduceMean operation (#6994)
This is a future replacement of `Reduce` operation.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 05:39:06 +0000 (14:39 +0900)]
[nnpkg_tool] Introduce nnpkg_test.sh (#7020)
Introduce a tool for running an nnpackage testcase.
It will run nnpackage with nnpackage_run, and compare the result.
For usage, find README.md.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Thu, 29 Aug 2019 05:35:18 +0000 (14:35 +0900)]
[neurun] Apply InternalBufferManager to RNN on acl_cl (#6978)
Apply InternalBufferManager to RNN on acl_cl
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 05:26:09 +0000 (14:26 +0900)]
[moco-tf] Introduce ProgressReporter (#7013)
* [moco-tf] Introduce PhaseEventListener
This will introduce PhaseEventListener that implements logo::PhaseEventListener that dumps Phase event logs
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* rename
이상규/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 04:52:16 +0000 (13:52 +0900)]
Introduce model2nnpackage (#6976)
This patch introduces model2nnpackage.
It will take model file (either circle or tflite) and generate
containing nnpackage.
Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 04:41:13 +0000 (13:41 +0900)]
[exo-tflite] Placeholders for TFLNodes.lst (#7019)
This will add TODO comments in TFLNodes.lst as placeholders for new TFL nodes to come
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 29 Aug 2019 03:48:56 +0000 (12:48 +0900)]
[exo-tflite] Use throw instead of assert (#7006)
The current implementation of translateLocoTypeToTFLite may return an
invalid value if the code is built with release configuration.
This commit replaces assert with throw to prevent undefined behavior.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 29 Aug 2019 03:48:35 +0000 (12:48 +0900)]
[moco-log] Extract LoggingContext class (#6999)
LoggingContext class is currently declared in Log.h, and thus it is
impossible to access this class declaration without including macros
defined in Log.h, which leads to macro name conflicts in some case.
This commit introduce a dedicated file for LoggingContext class to
allow users to access LoggingContext class without the risk of macro
name conflicts.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박종현/On-Device Lab(SR)/Staff Engineer/삼성전자 [Thu, 29 Aug 2019 01:09:26 +0000 (10:09 +0900)]
[nnkit] Remove empty directories (#6967)
All the components are now at the top-level.
Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 00:38:47 +0000 (09:38 +0900)]
[moco-tf] Use make_shape_inference_data in fix_shape (#6998)
This will use make_shape_inference_data() for patterns of make_unique() + copy_shape_values() with ShapeInferenceData
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 00:30:55 +0000 (09:30 +0900)]
[exo-tflite] Enabling OperationExporter to handle TFLNode (#7001)
This enables OperationExporter to handle TFLNode
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
박세희/On-Device Lab(SR)/Principal Engineer/삼성전자 [Thu, 29 Aug 2019 00:30:02 +0000 (09:30 +0900)]
[moco-tf] Use logo::Pass for Transform (#7000)
This will make moco::tf to use logo::Pass for Transform class as for first step of migration
Signed-off-by: SaeHie Park <saehie.park@samsung.com>
[nnc] Do not set / check names of operations (#6981)
Operation names are going to be removed.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir2loco] Switch to tensor names (#6974)
Do not set operation names; set tensor names instead.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_onnx] Do not set names of operations (#6956)
Remove useless setting of operation names.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
윤현식/On-Device Lab(SR)/Principal Engineer/삼성전자 [Wed, 28 Aug 2019 10:56:13 +0000 (19:56 +0900)]
[exo-tflite] Macros to wrap assert and std::runtime_error (#6919)
* [exo-tflite] Macros to wrap assert and std::runtime_error
Two macros were added that wrap assert and std::runtime_error
Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* renamed
김용섭/On-Device Lab(SR)/Engineer/삼성전자 [Wed, 28 Aug 2019 08:21:09 +0000 (17:21 +0900)]
[neurun] Recover prepare of IFunction (#6975)
Recover prepare of IFunction for #6156
Signed-off-by: Yongseop Kim <yons.kim@samsung.com>
[mir_tflite] Do not set names of operations (#6955)
Remove useless setting of operation names.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_caffe] Do not set names of operations (#6953)
Remove useless setting of operation names.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_onnx] Register Transpose operator converter (#6947)
It was lost not long ago after improper resolving of a merge conflict.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_onnx] Add support for auto_pad attribute (#6943)
Add support for `auto_pad` attribute to `Conv`, `AveragePool`, `MaxPool` operator converters.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>
[mir_caffe2] Do not set names of operations (#6812)
Remove useless setting of operation names.
Signed-off-by: Sergei Barannikov <s.barannikov@samsung.com>