0a1cad52ab
Avoid Constant data copy inside Reshape constant folding ( #6410 )
...
* Avoid Constant data copy inside Reshape constant folding
* Fix Codestyle
* Updated Squeeze, Unsqueeze cf
* Deprecate set_data_shape method
* Fix Pruning
2021-07-29 17:51:25 +03:00
e9011a9536
ShapeOf Sub-Graphs Simplification ( #6308 )
...
* ShapeOf Sub-Graphs Simplification
* Removed graph visualization
* Review comment
* comments resolved
2021-07-07 18:12:14 +03:00
d76d267478
Add support for fp64 in Convert's evaluate ( #5911 )
...
It's required for t2t-vit models.
2021-06-16 13:46:33 +03:00
dc0d482c23
Add support for fp64 in Sqrt's evaluate method ( #5913 )
...
It's required for t2t-vit models.
2021-06-16 13:45:36 +03:00
6d2740a335
[nGraph] Gather 7 evaluate ( #5088 )
...
* initial working solution
* constant_folding for Gather-7 successfully enabled
* successfuuly added remaining unittests
* removed redundant evaluate from private
* fixed evaluate for dynamic axis
* fix pre-commit for KMB: added default batch_dims=0 to runtime/reference
* clang_format_fix
* added include ngraph/shape.hpp to fix compilation for onednn-plugin
* temporary removed out of bound check for gather
* removed optional argument for batch_dims
* added const
* finally successfully run on ARM: removed redundant declaration
* style fix
* changed argument types from size_t -> int64_t
* changed back to size_t
2021-04-12 17:01:39 +03:00
a748c26fee
Align copyright notice in cpp and cmake source files (CVS-51320) ( #4950 )
2021-03-25 02:40:09 +03:00
c9d0292929
Add DilatedConvolutionConverter, BatchToSpaceFusion and SpaceToBatchF… ( #4689 )
2021-03-24 16:00:30 +01:00
ec5df02dac
Revise OneHot operation reference implementation ( #4330 )
...
* Auto benchmarks for evaluate(...) and extra unit tests.
* Test suite for manual i/o examination.
* Template on input type. Constant folding tests the 'count_ops_of_type' issue.
* Multidim manual tests
* NgOp&Ref: Unit tests Ok. NB evaluates_map
* Stable. Negative indices as if larger than depth. Benchmarks removed.
* Added Serialization SLT
* I/O shapes consistency checks
* Consistency checks in evaluate(). Removed the ref from evaluates_map.
* removed excessive #include from evaluates_map
* Code style
2021-03-16 11:28:30 +03:00
f1ffe4f9fd
nGraph value propagation ( #3368 )
...
* partial value and shape propagation
* TopK k propagation
* stylish commit
* style again
* Precalc Folding and broadcasting fix
* style
* Self review
* forgetting the unused values
* deleting unused variables
* Review
* self review
* disaster mistake fixed
* tests corrected
* mod to be inferred while creation
* LPT test fix
* tests fixes
* fix
* tests
* dynamic backend fix
* comments
* disable propagation
2021-02-02 12:03:56 +03:00
303c549791
nGraph copyright update for 2021 ( #3832 )
2021-01-13 16:43:04 +03:00
1ac3caf472
fix constant folding for sub graph ops ( #3534 )
2020-12-10 11:56:16 +03:00
256e047ad2
Revert "Deprecate global element types ( #3444 )" ( #3468 )
...
* Revert "Deprecate global element types (#3444 )"
This reverts commit 071fb9d1c6
2020-12-04 13:28:53 +03:00
6467c64000
Remove opset0 support and undesired passes from Interpreter backend ( #1469 )
...
* Move evaluate() interface from some OPs to Interpreter
* commit
* Move shuffle channels reference to OP's evaluate
* Add some operations missed in evaluate_node
* Fix select references invocation from evaluate_node()
* Activation refs (#2 )
* HardSigmoid
* Elu
* Selu
* Gelu
* Move to test runtime
* Rollback donwgrade passes delition
* Initial batch to space refs
* Return opset1_upgrade
* WIP: Add space to batch evaluate
* Fix space to batch
* add evaluates function in evaluates_map (#4 )
* Add space to batch evaluate
* Fix crop in batch to space references
* Remove vectors reallocation in evaluates for b2s and s2b
* .
* Add SpaceToDepth evaluate
* Add depth to space evaluate
* Remove code duplication depth to space evaluate
* Fix some failed layer tests
* Ngraph test (#3 )
* Remove some v0 ops & fix some tests
* Fixes BatchNorm
* Next
* dd
* s
* Add dot & replace slice refs
* d
* dkj
* Review fixes part 1
* Fixes. Part 2
* Fixes. Part 3
* Enable cells refs in evaluate map
* Fix some failed layer tests
* Some more fixes
* Fix code style (#6 )
* Tests (#7 )
* PriorBox
* Mod
* NormilizeL2
* Update prior_box.hpp
* Fix one hot ref call
* .
* Select (#8 )
* Select
* Fix code style
* Fix select messages
* ReverseSeq (#9 )
* ReverseSeq
* Select
* ExtractImagePatches, Seqence
* Fix Code Style
* remove extra
* Remove etra line@
* Add fake quantize reference
* Align convolution layer tests instantiations with updated definition
* Disabled some failed LPT tests
* Disabled some failed LPT tests
* Remove undesired changes
* Update unit-test manifests + some code cleanup
* Fix code style (#10 )
* Normalize L2 refs support (from PR #2327 )
* Fix code style
* Apply review comments. Part 1 (#11 )
* Apply first part of review comments
* Update onnx_import.in.cpp
* Remove redundant reshape from shuffle_channels evaluate
* Decompose GroupConvolution
* [IE Ngraph] Fix some operation inheritance (#13 )
* [IE TESTS] Depth2Space
* Space2Depth
* ShuffleChannels
* Fix ode style
* Fix code style
* [IE NGraph] Remove decompose op (#14 )
* .
* Fix loosing control dependency in replace_node
* Fix loosing control dependency in replace_node
* Fix code style
* Fix FQ references build on windows
* Fix code style
* Apply comments (#15 )
* [Ie Ngraph] Remove using v1::Add
* [Ie Ngraph] Remove using v1::Mutliply
* [Ie Ngraph] Remove using v1::Subtract
* [Ie Ngraph] Remove using v1::Divide
* [Ie Ngraph] Remove using v1::Equal
* [Ie Ngraph] Remove using v1::Greater
* [Ie Ngraph] Remove using v1::Greater_eq
* [Ie Ngraph] Remove using v1::Less
* [Ie Ngraph] Remove using v1::LessEq
* [Ie Ngraph] Remove using operator+
* [Ie Ngraph] Remove using operator/
* [Ie Ngraph] Remove using operator*
* [Ie Ngraph] Remove using operator-
* Fix code style
* Ci (#16 )
* Fix CentOS compilation
* Revert ngraph::op::vo::Multiply removing due to OpenCV
* Android fix (#17 )
* fix failures
* Fix code style
* Add (#18 )
* Android fix
* Add
* Add in opset1 upgrade pass
* Add in opset1 upgrade pass
* Remove v0::Add, Reverted removing v0::Multiply (#19 )
* Remove overloaded math operators from PyNgraph
* Remove overloaded math operators from PyNgraph
* Fix gna tests (#20 )
* Fix gna tests
* Squashed commit of the following:
commit 565b504c1c#2570 )
* GitHub CI: Add files_size.yml
* Update job name
commit ab0fb29853#2628 )
* C++ feature in C `typedef struct` code.
* The warning can be promoted to error in dependent projects.
C5208: unnamed class used in typedef name cannot declare members other than
non-static data members, member enumerations, or member classes
commit 15a338e89b#2502 )
It allows skipping inference-engine/thirdparty/pugixml and using the
system copy instead.
Thanks to @Osse for helping understand cmake scoping rules.
Co-authored-by: Helmut Grohne <helmut.grohne@intenta.de>
commit 7ac8cd85863a2e33962c5835974fad#21 )
* Disable failed tests on ia32
* Remove redundant broadcast from MVN ref
* Fix missed GatherND in opset_int_tbl + code style
* Remove one extra temporary buffer from MVN ref
* Merge master (#22 )
* Leaky relu transformation refactor (#2640 )
* Refactored LeakyRelu transformation
* Added unit test for LeakyRelu transformation + removed duplicate test function valued_const
* nGraph implementation of NMS-5 (without `evaluate()`) (#2651 )
* Written nGraph NMS-5 without evaluate().
* Used NGRAPH_RTTI_DECLARATION.
* setupvars.sh: Updated setting pyenv error to warning. (#2663 )
* Fix itt build (#2662 )
* Loop-5 operation specification (#2291 )
The Loop-5 operation specification
* Time tests improvements (#2642 )
* Remove extra functions from run_timetest.py
* Add `log.debug` of raw and aggregated statistics in run_timetest.py
* Implement storing of models locally for test_timetest.py
* Fixed CVS-35316 (#2072 )
* Extend MO for operation GatherND (#2540 )
* Extend MO for operation GatherND
* Update documentation
* Rename GatherNd.py to gathernd.py
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Add hsigmoid op to ngraph (#2647 )
* [IE CLDNN] Fixes for GatherTree and ReverseSequence (#2660 )
* ReorgYolo reference implementation (#2384 )
* Align ReorgYolo to the spec (vector strides -> int stride)
* ReorgYolo ref impl
* ReorgYolo evaluate method
* ReorgYolo tests
* Tests update
* Style apply
* Add some coments
* Code refactor
* Comment update
* Style apply
* Build fix, mark evaluate as override
* Revert "Align ReorgYolo to the spec (vector strides -> int stride)"
* Use int_executable instead of evaluate
* Use char* instead of templates
* Code refactor
* Comment update
* Code review comment
* Add constructor aligned with spec
* Update shape validation
* Update attributes tests
* Add type_prop tests
* Update backend tests
* Add single layer tests
* Update the spec
* Remove wrong transformation test
* Add ReorgYolo to evaluates_map
* code style
Co-authored-by: Evgeny Lazarev <evgeny.lazarev@intel.com>
Co-authored-by: Vladimir Gavrilov <vladimir.gavrilov@intel.com>
Co-authored-by: Artyom Anokhov <artyom.anokhov@intel.com>
Co-authored-by: Andrey Somsikov <andrey.somsikov@intel.com>
Co-authored-by: Vitaliy Urusovskij <vitaliy.urusovskij@intel.com>
Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
Co-authored-by: Roman Kazantsev <roman.kazantsev@intel.com>
Co-authored-by: iliya mironov <iliya.mironov@intel.com>
Co-authored-by: Vladimir Paramuzov <vladimir.paramuzov@intel.com>
Co-authored-by: Katarzyna Mitrus <katarzyna.mitrus@intel.com>
* RegionYolo
* Apply review comments
* Merge remote-tracking branch 'upstream/master' into update_evaluates
# Conflicts:
# ngraph/core/src/op/mvn.cpp
# ngraph/test/backend/fused_op.in.cpp
# ngraph/test/runtime/ie/unit_test.manifest
# ngraph/test/runtime/interpreter/int_executable.hpp
# ngraph/test/runtime/interpreter/opset_int_tbl.hpp
# ngraph/test/runtime/interpreter/unit_test.manifest
# ngraph/test/runtime/opset0_tbl.hpp
* Apply code style
* Apply comments
* Apply code style
* Fix RegionYolo evaluate redefinition
* Removed defines from evaluates map
* Apply code style
* Fix MVN ref
* rename select reference argument
* Fix code style
* Fix Fake Quantize references calculation (#24 )
* Fix MVN ref
* Fix MVN & adding NMS
* Fix TI
* Temporary relax comparison threshold for FQ SLT
* Fix GPU LPT Tests
* Add explicit rounding mode seetting in FQ references
* Apply code style
* Rollback op_is test deletion
* Apply code style
* Fix merge conflict resolving issues
* Apply code style
Co-authored-by: Irina Efode <irina.efode@intel.com>
Co-authored-by: Anton Zaytsev <anton.zaytsev@intel.com>
Co-authored-by: Evgeny Lazarev <evgeny.lazarev@intel.com>
Co-authored-by: Vladimir Gavrilov <vladimir.gavrilov@intel.com>
Co-authored-by: Artyom Anokhov <artyom.anokhov@intel.com>
Co-authored-by: Andrey Somsikov <andrey.somsikov@intel.com>
Co-authored-by: Vitaliy Urusovskij <vitaliy.urusovskij@intel.com>
Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
Co-authored-by: Roman Kazantsev <roman.kazantsev@intel.com>
Co-authored-by: iliya mironov <iliya.mironov@intel.com>
Co-authored-by: Vladimir Paramuzov <vladimir.paramuzov@intel.com>
Co-authored-by: Katarzyna Mitrus <katarzyna.mitrus@intel.com>
2020-12-03 12:36:34 +03:00
071fb9d1c6
Deprecate global element types ( #3444 )
...
* Removed global types
* Fixed ONNX importer and nGraph tests
* Fixed code style
* Disable some warnings for Windows
* Try to fix python build
* Extend deprecation message
* Fixed code style
* Fixed comments
2020-12-02 21:27:17 +03:00
6899a95e1c
ConstantFolding as a FunctionPass instead of GraphRewrite ( #3065 )
...
* Redundant op::Max CF removal
* Redundant op::Min CF removal
* Redundant op::Sum & op::Product CF removal
* CF Min and Max using evaluate()
* Arithmetic reduction CF pass removal
* Quantize op CF pass removal
* Convert op CF pass removal
* Logical reduction CF pass removal
* Select op CF pass removal
* OneHot CF pass removal
* Code formatting
* ScatterElements CF pass removal
* Gather CF pass removal
* Disable a Quantize op test that fails in CI
* CF pass cleanup
* Convert op cleanup and test adaptation to spec
* Possible fix for failing VPU tests
* Limit the types used in OneHot::evaluate
* Quantize op evaluator removal
* Refactor of Gather evaluator
* CF pass cleanup and adaptation to FunctionPass interface
* New CF pass implementation
* Fix the Tile::evaluate method for dynamic shapes
* Node shapes revalidation in CF pass
* Obsolete code cleanup
* Obsolete include removal
* Recursively fold subgraph nodes
* Obsolete include removal
* Revalidate each node in CF
* PR feedback
* Missing RTTI symbol definition
2020-11-20 20:36:05 +03:00
10b18a00c6
Improve reshapeability of models with eltwise nodes influencing shapes ( #2767 )
...
* Fix ElementwiseInputReshape transformation
Reshape node always needs to be inserted
in order to preserve ShapeOf nodes (reshapability of a model) that can potentially be above
elementwise node.
Refactor EltwiseInputReshape_test and EltwiseInputNormalization_test since the logic of maintaining reshape for eltwise has been changed.
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Merge EltwiseInputNormalization and EltwiseInputReshape transformations
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Remove Unsqueeze from Fused_op
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix code after code review #1
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix code after review #2
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix code review #4
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Perform full normalization based on shapes of all inputs to eltwise
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Refactor much to avoid old API and edges with unsqueeze_dims attribute
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix code after review
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
2020-11-16 09:50:41 +03:00
20df6eada6
Removal of obsolete constant folding passes ( #2902 )
...
* Redundant op::Max CF removal
* Redundant op::Min CF removal
* Redundant op::Sum & op::Product CF removal
* CF Min and Max using evaluate()
* Arithmetic reduction CF pass removal
* Quantize op CF pass removal
* Convert op CF pass removal
* Logical reduction CF pass removal
* Select op CF pass removal
* OneHot CF pass removal
* Code formatting
* ScatterElements CF pass removal
* Gather CF pass removal
* Disable a Quantize op test that fails in CI
* CF pass cleanup
* Convert op cleanup and test adaptation to spec
* Possible fix for failing VPU tests
* Limit the types used in OneHot::evaluate
* Quantize op evaluator removal
* Refactor of Gather evaluator
2020-11-11 15:49:40 +03:00
0ee774eb92
Remove obsoleted v0::Sum op ( #2938 )
2020-11-05 09:13:12 +03:00
d7d60b401e
Remove obsoleted v0::Slice and v0::Split ops ( #2908 )
...
* Remove obsoleted v0::Slice op
* Remove deprecated v0::Split op
* Fix build_graph tests
2020-11-05 09:11:45 +03:00
3f5584534a
Remove obsoleted v0::Reshape operator ( #2878 )
...
* Remove obsoleted v0::Reshape operator
* Fix handling negative dims on 32 bit platforms
Change-Id: Ia18e20df9a79fd0b7b38c434bee7140599833952
* apply code format
Change-Id: I4ddfb9e7033cbc97e167922f6ee5873d5487e551
2020-11-03 07:01:34 +03:00
d3c2d90d7f
Remove obsoleted v0::Or operator ( #2857 )
...
* Remove obsoleted v0::Or operator
* apply style format
2020-11-02 06:49:03 +03:00
3f35746c45
Remove deprecated v0::Reverse op ( #2899 )
...
* Remove deprecated v0::Reverse op
* fix unit tests
* add include for std::inserter
2020-11-02 06:46:30 +03:00
d6a9ef3a8f
Remove obsoleted v0::Not operator ( #2846 )
...
* Remove obsoleted v0::Not operator
* restore tests from ngraph/test/backend
2020-10-30 07:33:26 +03:00
1144eefe4d
Remove obsoleted Min, Max operators ( #2832 )
2020-10-29 11:32:50 +03:00
186e00fa2a
Remove obsoleted v0::Product op ( #2860 )
2020-10-28 07:12:52 +03:00
73c40722fd
Remove obsoleted v0::Gather and v0::GatherND ( #2826 )
2020-10-27 10:07:48 +03:00
4d2674962d
Remove obsoleted v0::Broadcast and BroadcastLike operators ( #2779 )
...
* Remove obsoleted v0::Broadcast and BroadcastLike operators
* remove NGRAPH_DEPRECATED marks from autobroadcast functions
* restore NGRAPH_SUPPRESS_DEPRECATED_START in autobroadcast.cpp
2020-10-26 06:32:49 +03:00
32b886a892
Remove obsoleted Dequantize op ( #2780 )
...
* Remove obsoleted Dequantize op
* apply code style
2020-10-23 11:25:08 +03:00
83670dd5cb
Remove deprecated Any op from nGraph ( #2719 )
2020-10-20 12:36:46 +03:00
3f55733b43
Updated ConstantFolding pass to propagate friendly name ( #2572 )
...
* Fixed concat output name when decomposing shape_of in CF
* Added friendly_name propagation for CF pass
* Propagate runtime info inside ConstantFolding pass
* Fixed rt info merge strategy to avoid cases when merge is not implemented but attributes number is equal to 1
2020-10-08 09:16:28 +03:00
d25bb6314a
Deprecate nGraph v0 ops and builders ( #1856 )
...
* Deprecate nGraph v0 ops
* Fixed build
* Added deprecated defines to fix windows
2020-08-20 12:27:14 +03:00
a63c8d9537
Optimize Concat operation ( #1812 )
...
* 1d case optimization
* code refactor
* concat optimization
* removed using template for concat
* unit tests to concat constant folding
* synchro with current master
2020-08-18 18:28:57 +03:00
1cb1f13eaa
v1::Pad evaluator ( #1771 )
...
* v1::Pad reference implementation
* ut fix: pad_negative_exterior_1d
* ut fix: pad_negative_exterior_1d_check_limits & pad_edge_1d
* Code formatting
* ut fix: pad_edge_1d_top_neg & pad_edge_1d_top_neg_bigger_than_tensor
* More Pad UT fixes
* Pad UT fixes: REFLECT mode
* Fix all Pad UTs
* Switch Pad evaluation in INT backend
* Non-template solution to v1::Pad::evaluate
* Always create v1::Pad with 4 inputs
* VS compilation error fix
* Python test fix
* Remove the v0::Pad constant folding pass
* Some extra checks in v1::Pad evaluator
* Code formatting
* Remove an obsolete CF test
2020-08-18 13:43:54 +03:00
aa35c59c5f
FGTT for CF ( #1792 )
...
* Added FGTT for ConstantFolding
* Added test for CF disabling
2020-08-17 15:39:58 +03:00
e4b411f027
Removed constant folding for Tile op ( #1595 )
2020-08-04 07:42:23 +03:00
d791d295aa
Use clone_function instead of specialize_function ( #1523 )
...
* Try to use clone_function instead of specialize_function
* Try to fix stress tests
* Remove redundant specialize_function
* Fixed TI clone
* Removed redundant code
* Uncomment threading tests
* Fixed docs
* copy function friendly name too
* Fixed copy rt_info
* Fixed comments
2020-08-03 18:23:02 +03:00
bd42f09e98
nGraph Transformations refactoring ( #931 )
...
This PR introduces next changes:
1. Transformations *_tbl.hpp files were replaced with direct registration in cpp files.
2. Plugins use pass::Manager to call conversion passes.
3. Transformations callback was moved to PassBase class as there is no more need to keep it in separate class
4. All pattern based transformations must be inherited from MatcherPass class. GraphRewrite class will be used only for matchers registration and execution on function.
MatcherPass class adds new features to pattern-based transformations approach:
* Allows to run matcher pass on a single node.
* Operations that were created inside transformation callback can be added to execution list to be available for pattern matching within single GraphRewrite.
5. GraphRewrite MatchClosure was replaced with MatcherPass. So all matchers will be registered as a MatcherPass.
6. Added pass::Manager::clear_state() method to avoid dependency with nodes that no longer belongs to function after replacement.
7. Some representative transformations were updated to use MatcherPass as an example.
8. Mul->Add sequence fusion transformation was replaced with LinOpSequenceFusion.
9. Pattern and callback registration code was moved to class c-tors (will be finished for remaining passes in other PR) .
10. Updated pass::Manager to get pass names only when NGRAPH_PROFILE_PASS_ENABLE enabled.
11. Moving towards removing PassProperty.
12. Added ngraph::pattern::wrap_type<T>(inputs, pred) to simplify pattern creation.
13. GraphRewrite was updated to execute MatcherPass more efficient.
2020-07-27 19:47:37 +03:00
29816f7a44
Remove get_arguments ( #1323 )
...
* Removed get_arguments
* Fixed code style
2020-07-17 09:50:06 +03:00
71d41a992f
Removed all and allreduce layers ( #1182 )
...
* Removed all and allreduce layers
* Removed reference implementations
2020-07-08 04:16:02 +03:00
59579eb437
Removed And operation ( #1183 )
2020-07-07 13:32:35 +03:00
5d1c5ee6a9
Removed nGraph experimental operations and headers ( #1197 )
2020-07-07 07:27:17 +03:00
934e0c61eb
Removed reference implementations for some data types ( #1086 )
2020-06-24 12:44:19 +03:00
f30dcc218c
publish master branch snapshot, revision 9df5eb1f84e13a35720a918f88324561222ab114
2020-05-20 01:13:06 +03:00
Gleb Kazantaev