[exo-tflite] Use link helper (#6200)

This commit revises exo-tflite to use link helper instead of
deprecated "node" method in GraphInput/GraphOutput.

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

[moco-tf] Canonicalize TFRelu6 (#6193)

This will add Relu6Canonicalizier that transforms TFRelu6 node to canonical ReLU6 node

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

Change the way getting layout of node in PermutationInsertionPass (#6148)

This commit changes the way getting layout of node in PermuttationInsertionPass from model::Subgraph to operation::LowerInfo.

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

[loco] Implement FeatureCodec shape inference (#6191)

This commit extends CanonicalShaapeInferenRule to accept
FeatureEncode and FeatureDecode nodes.

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

Remove softlink directory (#6188)

Remove softlink directory scripts/standalone - made for CI infra workaround
Update format check script

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

[moco-tf] Import as TFRelu (#6182)

This will change import Relu node as TFRelu from the knob

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

[loco] Introduce "link" helper (#6190)

This commit introdues "link" helper which associates
GraphInput/GraphOutout with Pull/Push node, respectively.

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

Format check hpp extension (#6187)

Add format check extension: hpp file
Fix format: nnfw_api.hpp

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

[moco-tf] Import as TFRelu6 (#6183)

This will change import Relu6 node as TFRelu6 from the knob

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

[moco-tf] LogHelper to dump Pad and PadData (#6185)

This will introduce LogHelper to dump Pad and PadData values

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

[tflite_benchmark] Insert pause between iterations (#6189)

tflite_benchmark now allows users to insert pause between iterations via
"PAUSE" environment variable.

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

[moco-tf] Revise COpCall test (#6184)

This will revisit CopCall node test to remove relation with graph connections

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

[enco] Fix copyright year (#6174)

This commit fixes copyright year for 2019 generated file

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

[loco] Implement Relu shape inference (#6161)

* [loco] Implement Relu shape inference

CanonicalShapeInferenceRule now properly infers the shape of relu node.

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

Update scheduled deprecated API. (#6163)

* Change old API `gfile.FastGFile` to `tf.gfile.GFile`

Signed-off-by: Sung-Jae Lee <sj925.lee@samsung.com>

Use external volume mount in standalone scripts (#6136)

It enables using external volume mount in CI test infra
- Reduce external code download overhead
- Remove pre-built acl binary dependency (especially on version up)

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

[moco-tf] Summary for TFSqueeze (#6162)

* [moco-tf] Summary for TFSqueeze

This commit introduces log summary for TFSqueeze

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Make use of pepper::str

[pepper-strcast] Fix missing copyright boilerplate (#6171)

This commit fixes missing copyright boilerplate

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

[moco-log] Fix missing copyright boilerplate (#6172)

This commit fixes missing copyright boilerplate

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

[locop] Fix copyright year (#6173)

This commit fixes copyright year for 2019 generated file

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

[locop] Extract CanonicalNodeSummaryBuilder (#6169)

This commit extracts CanonicalNodeSummaryBuilder from FormattedGraph
module, and put them in a header/source.

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

[moco-tf] implementing make_graph_builder_registry(..) function (#6175)

This adds `make_graph_builder_registry(..)` function, which makes graph builder registry that finds custom op graph builder and then searches into the default registry.

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

[locoex] TypeInferenceRule for custom op (#6170)

TypeInferenceRule and a test for custom op was committed.

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

[moco-tf] Knobs to import as TFRelu and TFRelu6 (#6181)

This will add two knobs to import as TFRelu and TFRelu6

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

[mir/Caffe2 importer] Improve support for constant tensors (#6176)

Rework support for constant tensors to not (almost) depend on concrete models.

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

[mocotest-tf] Enable basic Reshape test (#6164)

This commit enables basic Reshape test, Reshape_000.

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

[mocotest-tf] Enable test for dynamic reshape with wildcard dim (#6166)

This commit enables moco test which can check resolving dynamic reshape
with wildcard dimension.

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

[exo-tflite] Use loco shape inference framework (#6144)

This commit allows exo-tflite to use the shape inference framework
implemented in loco instead of exo-tflite internal shape inference
framework.

Please note that this feature is turned off by default.

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

[moco-tf] Introduce fmt helper (#6149)

This commit introduces fmt helper specialized for moco-tf.

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

[moco-tf] GraphBuilder for custom op (#6139)

* [moco-tf] GraphBuilder for custom op

This adds GraphBuilder, transforms, and a test for custom op.

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

[moco-tf] TFSqueeze shape inference (#6117)

* [moco-tf] TFSqueeze shape inference

This commit introduces shape inference for TFSqueeze

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Review fix: use set, use lambda for assertion

[moco-tf] Enable ResolveReshapeWildcardDim transform (#6154)

This commit registers and enables ResolveReshapeWildcardDim transform.

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

[locop] Extract NodeSummaryBuilder (#6150)

This commit introduces a new header file dedicated to
NodeSummaryBuilder and NodeSummaryBuilderFactory classes.

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

[moco-tf] Enable TFReshape canonicalizer (#6151)

This commit registers TFReshape canonicalizer to enable it.

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

[mir2loco] Support BiasAdd operation (#5901)

* Transform `mir::BiasAddOp` to `loco::BiasAdd<Domain::Tensor>` using `loco::BiasEncode>
* Test for this

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>

[moco-tf] Fix seg fault with LOG enabled (#6155)

This will fix segment fault when run with LOG enabled

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

[neurun] Use shape inference in interpreter (#6012)

* Use shape inference in interpreter

To support dynamic shape, use shape inference: convolution, maxpool, averagepool

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Fix unspecified shape checking and shape inference for depthwise conv

[moco-tf] Introduce TFRelu and TFRelu6 IR (#6143)

This will introduce TensorFlow dialect TFRelu and TFRelu6 IR and minimum required codes

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

[mir] Remove stale serialization support (#6120)

It is stale and were never used.

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

Fix a typo in PluginClassesAllocator for acl_neon (#6142)

This commit fixes a typo in PluginClassesAllocator for acl_neon.

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

[neuurn] Use mutex on ExecutorBase (#6131)

* Use mutex on ExecutorBase

Use mutex on ExecutorBase for thread-safe executor
Enable multi-thread execution test

Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
* Use std::lock_guard

[locop] Use default_node_desc directly (#6141)

The current implementation of BuiltinNodeSummaryBuilder invokes
default_node_desc through node_desc helper.

This commit removes this unnecessary redirection.

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

[moco-tf] Reshape canonicalize support ConstGen shape input (#6115)

* [moco-tf] Reshape canonicalize support ConstGen shape input

In canonicalization of TFReshape, shape input of Reshape supports not
only TF-dialect TFConst, but also canonical ConstGen. This is to
eliminate dependency on canonicalization order.

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

* Fix typo

* Remove ambiguity on shape()

[loco] Add CanonicalShapeInferenceRule (#6126)

This commit adds CanonicalShapeInferenceRule class with minimal
implementation.

The current implementation supports shape inference only for Push and
Pull nodes.

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

[moco-tf] Removing CompilerArgs.h (#6133)

With this commit, CompilerArgs.h was removed and header inclusion errors were fixed.

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

[locop] Show Push/Pull index field (#6132)

CanonicalNodeDescBuilder now shows the index attribute of Push/Pull
nodes.

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

[moco-tf] Logging compile stage (#6116)

This commit introduce logging onset of each compile stage.

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

[loco] Automatic Update Graph Input/Ouput Index (#6129)

This commit revises "GraphInput" and "GraphOuput" class to automatically
feed its index to "Pull" and "Push" nodes on node method call.

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

Fix verify_downloaded_file usage in run_test.sh (#6070)

run_test.sh downloads model files even they exist. It is fixed now

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

[moco-tf] Fix indentation of MaxPool test (#6125)

This will fix TODO indentation for MaxPool test

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

[moco-tf] Fix indentation for AvgPool test (#6124)

This will fix TODO indentation for AvgPool test

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

Support TransposeConv op for acl_neon (#5786)

This commit supports TransposeConv op for acl_neon backend.
  - Introduce NETransposeConvLayer
  - Introduce CPPUpsampeEx
  - Introduce CPPUpsampeKernelEx
  - Apply NETransposeConvLayer for acl_neon backend
  - Enable nnapi tests for TransposeConv

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

[moco-tf] remove unused code in PlaceholderGraphBuilder (#6113)

Unused code (to process shape info entered by user for unknown shape) is removed.

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

Change nnfw component diagram (#6073)

* Change nnfw component diagram

- Remove core structure layer
- Add ML Frameworks in diagram

Signed-off-by: Efimov Alexander <a.efimov@samsung.com>

[moco-tf] Enable moco to print custom op log (#6114)

By adding `COpNodeSummaryBuilder`, moco can print custom op info.

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

[loco] Introduce ShapeInference infrastructure (#6102)

* [loco] Introduce ShapeInference infrastructure

This commit introduces dialect-agnostic ShapeInference infrastructure.

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

[mir] Support 0-dimensional tensors (#6122)

Partially fixes compilation error of TFLIte StyleTransfer nework.

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

[nnc] Support 0-dimensional tensors in x86-c++ backend (#6121)

Fixes compilation error of TFLIte StyleTransfer nework.

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

[mir2loco] Support Const operation (#5794)

* Support Const operation transformation
* Append test with Const op containing float values

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>

[mir] Add missing include (#6119)

Add missing include `mir/TensorVariant.h` to `ConstantOp.h`.

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

Support mir::ReshapeOp transformation to loco::Reshape<Fixed> (#5952)

* Added support to transformer of reshape operation
* Added test for Reshape

Signed-off-by: Pavel Iliutchenko <p.iliutchenk@samsung.com>

[locoex] NodeSummaryBuilder for Custom op (#6097)

* [locoex] NodeSummaryBuilder for Custom op

This adds COpNodeSummaryBuilder to log custom op information.

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* printing opname to 'COp.Call'

* 1) using auto 2) using uint32_t instead of int

* summary() as private

* remove empty line. add assert

[loco] Introduce indexed method to Push/Pull (#6106)

It is currently impossible to check whether Push/Pull is indexed or not.

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

[moco-tf] Refactor fixed check function to template (#6108)

* [moco-tf] Refactor fixed check function to template

This commit refactors function for fixed reshape's shape input check.
New template function can accept TFConst and canonical ConstGen.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Return false on nullptr sourced

[mir-caffe2-importer] Perform operation sanity checks while creating (#6037)

Perform operation sanity checks while creating it

Signed-off-by: Ivan Vagin <ivan.vagin@samsung.com>

[mir] Fix number of outputs of `OutputOp` (#5993)

`OutputOp` is auxilary operation, which is used only to mark outputs of the Graph. It has one input and mistakenly had one output.

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

[tf2tflite] Enable Add_000 and Mul_000 (#6111)

This will enable Add_000 and Mul_000 basic test

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

[mocotest-tf] Enable Add_000 and Mul_000 (#6110)

This will enable Add_000 and Mul_000 basic test

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

[exo-tflite] Remove redundant export test (#6109)

tf2tflite is doing more organized and advanced test with value checking
and larger coverage, which makes this export test redundant.

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

[mocotest-tf] Adding a test for constant folding (#6104)

This adds a test case (TF_SMALL_NET_0011), which is a graph to test constant folding.

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

[locop] Support EltwiseAdd and EltwiseMul (#6098)

* [locop] Support EltwiseAdd and EltwiseMul

This will add locop to support EltwiseAdd and EltwiseMul to dump it's contents

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* remove todo msg

[locop] Remove comment for TensorConcat (#6103)

This will remove unnecessary TODO comment

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

[exo-tflite] Exporter for EltwiseAdd and EltwiseMul (#6100)

This will add exporter to support for EltwiseAdd and EltwiseMul

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

[loco] TypeForward for EltwiseAdd and EltwiseMul (#6101)

This will add TypeForward for EltwiseAdd and EltwiseMul node

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

[exo-tflite] Shapeinf for EltwiseAdd and EltwiseMul (#6099)

This will add to support ShapeInference for EltwiseAdd and EltwiseMul

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

Remove unused comment in BackendManager.cc (#6095)

A nit - remove unused code comment in BackendManager.cc
This has been left by accident.

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

[locomotiv] execution for elementwise mul (#6094)

* [locomotiv] execution for elementwise mul

This will introduce execution of EltwiseMul node

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* validate same domain

[moco-tf] Using composed GraphBuilderRegistry for custom op (#6093)

* [moco-tf] Using composed GraphBuilderRegistry for custom op

This is to use a composed GraphBuilderRegistry, instead of singleton GraphBuilderRegistry, to add GraphBuilder for custom op later.

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

[moco-tf] adding constant folding into optimizer (#6065)

This commit adds constant folding into optimizer and turns the Knob ON.

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

[moco-tf] Enable canonicalization for Add and Mul (#6096)

This will enable Canonicalization for TFAdd and TFMul node

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

[moco-tf] Import Mul node (#6086)

* [moco-tf] Import Mul node

This will enable import TensorFlow Mul node as TF dialect TFMul node

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

[moco-tf] Import Add node (#6083)

* [moco-tf] Import Add node

This will enable import TensorFlow Add node as TF dialect TFAdd node

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

[moco-tf] Introduce AddCanonicalizer (#6081)

This will introduce Canonicalizier for Add node. Supports only for Elementwise add for now.

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

[moco-tf] Import as TF dialect (#6091)

This will change import as TF dialects for AvgPool, MaxPool and Identity

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

[locop] Extract NodeSummary (#6076)

This commit introduces a dedicated header/source file for NodeSummary class.

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

[loco] Introduce NodeShape class (#6078)

* [loco] Introduce NodeShape class

This commit introduces NodeShape class which serves as a variant over
various Shape classes.

Signed-off-by: Jonghyun Park <jh1302.park@samsung.com>
* Fix bugs and add tests

[locomotiv] execution for elementwise add (#6079)

* [locomotiv] execution for elementwise add

This will introduce execution for EltwiseAdd node

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

* remove using namespace

* check if same domain not tensor

[moco-tf] Introduce MulCanonicalizer (#6082)

This will introduce Canonicalizer for Mul node. Supports only for Elementwise mul for now.

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

[moco-tf] Simplify TFFormattedGraph using pepper::str (#6080)

This commit simplifies TFFormattedGraph implementation using
"pepper::str" helper.

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

[moco-tf] use plier method (#6090)

This will fix to use plier method from old internal method

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

[moco-tf] Import as TFMaxPool with a knob (#6043)

* [moco-tf] Import as TFMaxPool with a knob

This will revise import to load MaxPool node as TFMaxPool with a knob

Signed-off-by: SaeHie Park <saehie.park@samsung.com>
* use plier and rename variable

[tf2tflite] Schema for additional info of Custom Op from user (#5939)

* [tf2tflite] Schema for additional info of Custom Op from user

This is a schema for additional info of Custom Op from user.
It was committed into _moco-tf_ but moved to here.
The name was renamed to CompileInfo.pbtxt (Previous name was Model.MetaInfo.pbtxt), which seems more friendlier. :-)

Signed-off-by: Hyun Sik Yoon <hyunsik.yoon@samsung.com>
* renaming CompileInfo to CustomInfo && move ./proto/CMakeFiles.txt to ./CMakeFiles.txt

* modify DT_INT32 to 15

[exo/tflite] prevent ceil_div function from overflow (#6088)

This commit prevents ceil_div function from overflow.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>

[moco-tf] Logging with TFNodeSummaryBuilderFactory (#6064)

TFNodeSummaryBuilderFactory is used to print all nodes (including unlinked TFNodes)

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

Introduce PermuteFactor into LowerInfo of operation (#6028)

This commit introduces PermuteFactor into LowerInfo of operation.

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

Refactor Backend to support sessions (#5894)

Backend had been only one object there cannot be more than one session.
This commit introduces `BackendContext` which depends on a specific
model and `Backend` object represents the backend itself(a singleton).

Resolve #5723

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

[moco/tf] Introduce DepthwiseConv2D Canonicalizer (#6047)

* [moco/tf] Introduce DepthwiseConv2D Canonicalizer

This commit introcude DepthwiseConv2D Canonicalizer to moco-tf.

Signed-off-by: seongwoo <sw4670.chae@samsung.com>
* apply comment.

* add codes for log.

[moco-tf] Fix build break (#6087)

This commit fixes build break related with API deisgn changes.

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

[moco-tf] Add a complex test case into ConstantFoldingTransform (#5936)

* [moco-tf] Add a complex test case into ConstantFoldingTransform

This commit adds a complex test case into ConstantFoldingTransform

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

[moco-tf] Transform to resolve wildcard dimesion of TFReshape (#6019)

* [moco-tf] Transform to resolve wildcard dimesion of TFReshape

This commit introduces ResolveReshapeWildcardDim transform, which tries
to resolve wildcard dimension of TFReshape.

Signed-off-by: Cheongyo Bahk <ch.bahk@samsung.com>
* Clarify by docs and function name