[tfl-verify] dummy driver (#8825)

This will introduce tfl-verify dummy driver from circle-verify

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

Make ncnn backend to support DepthwiseConv2D in row majar (#8826)

This commit makes backend to support DepthwiseConv2D in row major.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[moco] Introduce TFFakeQuantWithMinMaxVars (#8814)

This will introduce TFFakeQuantWithMinMaxVars node

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

[exo] Extract out FixedArityNode (#8810)

This commit extracts out FixedArityNode defined in TFLNodes.h. This is
mainly to use it for Circle dialect.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[exo] Fix typo (#8821)

This commit fixes typo for comment in TFLShapeInferenceRule.cpp

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[exo] Remove unused headers for Passes (#8822)

This commit removes unused 'loco/IR/Nodes.h' header from exo Passes

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[tf2tflite-dredd-pbtxt-test] adding test.lst (#8813)

This adds test.lst, which checks NET_0025

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>

[tfl-verify] Readme (#8817)

This will introduce tfl-verfiy project README file explaing what its purpose

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

Make ncnn backend to support Conv2D (#8811)

This commit makes ncnn backend to support Conv2D in row major.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[exo] Introduce Circle dialect (#8802)

This commit introduces structure for Circle dialect for loco ecosystem

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[nncc] Support BUILD_WHITELIST option (#8718)

* [nncc] Support BUILD_WHITELIST option

This commit allows users to restrict the modules to be built by setting
BUILD_WHITELIST option.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Update error message and add a relevant comment

* Use directory name

[tf2tflite-dredd-pb-test] Introducing tf2tflite-dredd-pb-test (#8779)

* [tf2tflite-dredd-pb-test] Introducing tf2tflite-dredd-pb-test

This adds tf2tflite-dredd-pb-test, which tests tflite file to check if optimization is done right, etc.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* remove tfkit

* remove downloading feature

* file existence checking macro

* remove tfkit again

* revising check_file_exist

[exo] Register TensorReduce conversion (#8809)

This commit registers conversion for TensorReduce to exo conversion
stage.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[res/TensorFlowTests] renaming rules (#8805)

This renames rules to [circle/tflite]_SDKver_type_req.rule.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>

[exo] Extract out FusedActFunc (#8798)

This commit extracts out enum class FusedActFunc. This is mainly to use
them in Circle dialect as well.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[nnpkg-test] Change nncc unzip path for test (#8804)

Change nncc install path into workspace

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

Make ncnn backend to support TransposeConv in row major (#8806)

This commit makes ncnn(srcn) backend to support TransposeConv in row major.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[runtime] Set installed libraries and binaries rpath (#8800)

Set rpath: same implementation with nncc

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

[neurun,ncnn-backend] Implement instance norm kernel for NHWC layout (#8795)

This commit implements instance norm kernel for NHWC layout
- Applying loop parallelism with OpenMP

Signed-off-by: JiHwan Yeo <jihwan.yeo@samsung.com>

Introduce HWIO and IOHW (#8785)

* Introduce a kernel layout as HWIO into srcn backend

This commit instroduces a kernel layout as HWIO into srcn backend.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
* Introduce a kernel layout as IOHW into srcn backend

This commit introduces a kernel layout as HWIO into srcn backend.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>
* Introduce getFilterPermutation into ncnn

This commit introduces getFilterPermutation into ncnn.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[tf2circle-dredd-pbtxt-test] adding rule files for NET_0025 and NET_0030 (#8786)

* [tf2circle-dredd-pbtxt-test] adding rule files for NET_0025 and NET_0030

This adds two rule files in for NET_0025 and NET_0030. This will be used to test circle files.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* fix rule name

* remove commnent

* Update res/TensorFlowTests/NET_0025/test.circle_rule

OMG. Thanks. :-)

Co-Authored-By: hyunsik-yoon <hyunsik.yoon@samsung.com>

Enable nnpackage_run on gbs build (#8801)

- Add build deps into nnfw.spec file

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>

[exo] Convert TensorReduce (#8741)

* [exo] Convert TensorReduce

This commit introduces convert stage for canonical TensorReduce node

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Strict check

[nnpkg_tools] gen_golden.py can run on tensorflow 2.0 (#8778)

`tf.contrib.framework.get_placeholders` is replaced with own implementation.
`gen_golden.py` can run on tensorflow 2.0.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>

[CI/script] Rename nnpackage testsuite generation script (#8773)

Rename nnpackage testsuite generation script build_nnpkg.sh to test_make_nnpkg.sh

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

[neurun] Shared neurun core library (#8777)

Change neurun_core to shared library - libneurun_core.so
When neurun_core is static, core is included to libneuralnetworks.so and libnnfw-dev.so and it make duplication

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

[exo] passing param of InputHandler as reference type (#8768)

vector is now passed as a reference type.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>

[android] Remove the old RootFS builder script (#8782)

This commit removes `build_android_rootfs.sh` which has been replaced by
`build_android_ndk.sh`.

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

Modify convertCoordinates of ncnn to support various filter layouts. (#8784)

This commit modifies convertCoordinates of ncnn to support various filter layouts.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[verify-package] Add LAYOUT_191105 check (#8781)

Let's check whether the package has a specific layout.

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

Apply backend layout for getTensorDesc() of ncnn (#8755)

This commit applies backend layout for getTensorDesc() of ncnn.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[exo] Add missing argument of static_assert (#8760)

This commit will add missing argument of `static_assert` in `exo/src/GraphBlock.h`

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

Move registering tensor info of ncnn into ShapeFixer (#8783)

This commit moves registering tensor info of ncnn into ShapeFixer to resolve special case.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[neurun] Fix warning from Android build (#8787)

This commit fixes narrowing conversion warning from Android compiler.

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

Introduce to set layout of io tensors into nnfw api (#8715)

This commit introduces to set layout of io tensors into nnfw api.
  - Introduce NNFW_LAYOUT into nnfw_api
  - Introduce nnfw_set_input_layout and nnfw_set_output_layout into nnfw api

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

Remove LowerInfoMap from IKernelGenerator (#8780)

This commit removes LowerInfoMap from IKernelGenerator.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[moco] Fix comment for input type (#8770)

This will fix comment of file name for supporting types of graph input

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

[doc] Update xu4 tizen setting document (#8757)

Update xu4 tizen setting document for latest environment (tested on tizen 5.5)
- Merge flashing boot and image
- Remove expand rfs: already expanded on /opt/usr

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

Add LowerInfoMap and frontend_layout into ShapeFixer of srcn (#8776)

This commit adds LowerInfoMap and frontend_layout as members into ShapeFixer of srcn.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

Introduce asKernelShape into Convert of ncnn (#8772)

This commit introduces asKernelShape into Convert of ncnn for converting operand of weights.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

Add copyright for add operation (#8771)

Added copyright since binaryop.{h,cpp} are modified.

Signed-off-by: JiHwan Yeo <jihwan.yeo@samsung.com>

[compute,ncnn] Add copyright and licence for instance_norm (#8765)

I added copyright and license since instance_norm.{h,cpp} are modified.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>

Loader support PReLU operation (#8769)

- Add PReLU into base_loader.h

Signed-off-by: Chunseok Lee <chunseok.lee@samsung.com>

[nnpackge_tools] introduce gen_golden.py (#8719)

It introduces `gen_golden.py` tool.
You can generate golden data with

```
$ gen_golden.py -o out_dir /your/path/to/pb
```

Then, you will get out_dir/{input,expected}.h5.

No need to specify inputs and outputs.
No need to build nncc.

Prerequisite: tensorflow running environment.

Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>

[tf2tflite-dredd-pbtxt-test] introducing rule-lib.sh (#8761)

* [tf2tflite-dredd-pbtxt-test] introducing rule-lib.sh

rule-lib.sh is a library of shell script functions that are used by dredd rule.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* Update compiler/tf2tflite-dredd-pbtxt-test/rule-lib.sh

Co-Authored-By: hyunsik-yoon <hyunsik.yoon@samsung.com>

[res/TensorFlowTests] Adding dredd rule for NET_0025 (#8763)

This adds dredd rule into NET_0025.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>

[exo] Adding assert (#8762)

This adds `assert` to check parameter.
for https://github.sec.samsung.net/STAR/nnfw/pull/8752#discussion_r295727

Signed-off-by: Hyun Sik Yoon hyunsik.yoon@samsung.com

[moco] Remove temporary graph output shape settings (#8767)

This will remove temporary graph output shape settings in Importer as it is now set in moco-tf frontend

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

Add TF2CIRCLE_EXIST and TF2CIRCLE_RUNNABLE check (#8754)

From now on, "nnas verify-package" command is able to verify whether
there is tf2circle (TF2CIRCLE_EXIST), and it has no missing dependencies
(TF2CIRCLE_RUNNABLE).

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

[neurun] Support SpaceToBatch from BaseLoader (#8727)

Support SpaceToBatch operation from frontend BaseLoader.

Draft PR: #8404

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

Add LowerInfoMap as member into class IKernelGenerator (#8742)

This commit LowerInfoMap as member into class IKernelGenerator.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

 [ncnn-backend] Support instance norm operation for NCHW layout  (#8721)

NCNN instance norm kernel porting
- Write updated values to output tensor instead of input tensor
- This commit supports NCHW layout only
  - Later PR will support instance norm for NHWC layout

Signed-off-by: JiHwan Yeo <jihwan.yeo@samsung.com>

[nnpkg-test] Update tf2tflite test name (#8759)

Test is renamed: tf2tflite_value_pbtxt_test
Test path: compiler/tf2tflite-value-pbtxt-test

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

[exo] Type & shape inference for TFLMean (#8739)

This commit introduces type and shape inference for TFLMean

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[tf2tflite-dredd-pbtxt-test] Adding CMakeLists.txt and runner.sh (#8744)

* [tf2tflite-dredd-pbtxt-test] Adding CMakeLists.txt and runner.sh

CMakeLists.txt and runner.sh files were added.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* modify comment

* rules -> rule

* TARGET_FILENAME -> TARGET_TESTNAME

[exo] Refactoring EltwiseBinaryConverter using DomainConverter (#8752)

This refactors EltwiseBinaryConverter using DomainConverter (and input handler).

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>

[neurun/cker] Add Elementwise sub support in cpu (#8554)

* [neurun/cker] Add Elementwise sub support in cpu

Added Sub operation to neurun cpu backend and kernel to cker.
It is copied from Add and renamed -i add to sub and changed
+ to -
Related issue: 7814

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
* Remove sub from framework test since just broadcast is tests

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
* Templatize Sub and Add for float

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
* Fix rebase issue because of renaming runtimes to runtime

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>

[neurun/cker] Add Logistic operation support in cpu (#8572)

* [neurun/cker]Add Logistic operation support in cpu

Added support of logistic operation (which is sigmoid)

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>
* Address feedback: use std::exp instead of exp

Signed-off-by: Poshshoev Dilshodzhon <d.poshshoev@samsung.com>

[neurun] Print error when a test was not found (#8728)

Print error when a framework test was not found. It is printed in file
`framework_test.log`.

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

[moco] Set graph input shape in Importer (#8746)

This will enable graph input shape setting after loading in Importer

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

Remove oneco-integration-test (#8747)

This commit removes oneco-integration-test, which is replaced by
oneco-value-pbtxt-test

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

Remove tf2tflite-integration-test (#8748)

This commit removes tf2tflite-integration-test, which is replaced by
tf2tflite-value-pbtxt-test.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

Remove tf2tflite-model-test (#8749)

This commit removes tf2tflite-model-test, which is replaced by
tf2tflite-value-pb-test

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

Remove moco-integration-test (#8750)

This commit removes moco-integration-test, which replaced by
moco-value-pbtxt-test

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[exo] Register Fuse Rsqrt and SquaredDifference (#8705)

This commit registers FuseRsqrtPass and FuseSquaredDifferencePass to exo
TFL optimize stage. Note that FuseSquaredDifferencePass is turn off by
default.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[nncc] Install tf2circle (#8725)

Let's include tf2circle in nncc package. Please note that tf2nnpkg
still uses tf2ftlite and tflite2circle instead of this tf2circle.

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

[exo] Export TFLMean (#8740)

This commit implements export stage of TFLMean for both tflite and
circle output

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[nncc] Show configure progress (#8679)

* [nncc] Show configure progress

From now on, running nncc configure shows configuration progress for
each individual components.

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

Introduce oneco-value-pbtxt-test (#8737)

This commit introduces oneco value test for pbtxt input model. It copies
existing oneco-integration-test only with renaming.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

Introduce moco-value-pbtxt-test (#8738)

This commit introduces moco value test for pbtxt input model. It copies
existing moco-integration-test only with renaming.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

Introduce tf2tflite-value-pb-test (#8735)

This commit introduces tf2tflite value test for pb input model. It copies
existing tf2tflite-model-test only with renaming.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

Introduce tf2tflite-value-pbtxt-test (#8736)

This commit introduces tf2tflite value test for pbtxt input model. It copies
existing tf2tflite-integration-test only with renaming.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[exo] introducing DomainConverter (#8745)

This adds DomainConverter and InputHandler, which are helpers used while converting a canonical node in Feature domain into a TFL node (Tensor domain).

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>

[moco] GraphOutputIndexQueryService in TF dialect (#8726)

This will add GraphOutputIndexQueryService in TF dialect

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

Enable ncnn(srcn) tensor to support NCHW (#8723)

This commit enables ncnn(srcn) tensor to support NCHW(row major).

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[tf2tflite-dredd-pbtxt-test] renaming tf2tflite-quality-test to tf2tflite-dredd-pbtxt-test (#8717)

* [tf2tflite-dredd-pbtxt-test] renaming tf2tflite-quality-test

As discussed offlie, tf2tflite-qualit-test is renamed to tf2tflite-dredd-pbtxt-test.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* removed compiler/tf2tflite-quality-test/.gitignore

[loader] Fix infinite recursive call (#8724)

Explicit call parent class's method

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

[moco] TFPlaceholder helpers for index query (#8720)

This will add two helper methods for GraphInputIndex query service

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

[neurun] Replace OperandIndex in DepthToSpace and SpaceToDepth Params (#8707)

Replace `OperandIndex` in `DepthToSpace::Param` and `SpaceToDepth::Param` with `int32_t`.

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

Introduce a configuration for layout of ncnn (#8621)

This commit introduces a configuration for layout of ncnn.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[loco] GraphInputIndexQueryService and input_nodes() (#8713)

This will introduce GraphInputIndexQueryService interface and input_nodes() method to gather Graph Input nodes

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

Make ExecutorBase supports NCHW frontend layout (#8692)

This commit makes ExecutorBase supports NCHW frontend layout.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[neurun] Replace OperandIndex in LocalResponseNormalization::Param (#8708)

Replace `OperandIndex` in `LocalResponseNormalization::Param` with `int32_t` / `float`.

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

[lib/ncnn] Fix depthwise conv NCHW kernel bug (#8711)

Fix depthwise convolution NCHW 3x3 weight same padding kernel bug

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>

[base_loader] Throw when file open failed (#8706)

Throw with a brief message when file open failed

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

[exo] TFLMean IR (#8698)

This commit introduces TFLMean IR

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[moco] Fix typo on TFMean shape inference (#8699)

This commit fixes typo on TFMean shape inference

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[nnpackage_run] Use HDF5 C++ interface (#8671)

Use HDF5 C++ interface instead of C.

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

Do NOT install locomotiv (#8690)

locomotiv is now built as a static library, and thus it is no longer
necessary to install this library.

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

[moco] Use moco Importer.h (#8702)

This will update to use moco/Importer.h in moco and remove old file in sub folder

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

[test] Fix misleading path (#8693)

Nit: fix path in `test_arm_neurun_acl_cl.sh`.

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

[neurun] Introduce InstanceNorm op (#8526)

This commit introduces InstanceNorm op into runtime frontend.
  - Introduce InstanceNormNode into neurun
  - Introduce InstanceNorm op into frontend
  - Use default value if value of epsilon is zero

Signed-off-by: JiHwan Yeo <jihwan.yeo@samsung.com>

[exo] Knob for fuse SquaredDifference (#8696)

This commit introduces UseFuseSquaredDifferencePass knob

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[exo] Register EltwiseSqrt Converter (#8697)

This commit registers EltwiseSqrt Converter to exo conversion stage

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[exo] Remove redundant steps of type and shape inference (#8695)

This removes redundant steps of type and shape inference.

Note that convert() and optimize() has already performed type and shape inference.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>

[angkor] Change num_elements behavior (#8684)

The current implementation treats rank-0 tensors as a special tensor
which has no elements.

This semantics has a conflict with the well-known standard which treats
a rank-0 tensor as a scalar.

This commit resolves this semantic conflict.

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

[exo] Fuse Squared Difference (#8654)

This commit introduces a pass to fuse Squared Difference pattern

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>

[moco-tf] Remove alias and use moco (#8688)

This will remove moco::tf alias for moco and update to use moco types

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

Introduce Readers for nchw (#8663)

This commit introduces Readers for nchw.

Signed-off-by: jiseob.jang <jiseob.jang@samsung.com>

[exo] Export Sqrt, Rsqrt, SquaredDifference (#8652)

This commit implements export stage for TFLSqrt, TFLRsqrt and
TFLSquaredDifference. Note that both TFLite exporter and Circle exporter
implemented.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>