IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
1,014 Commits 38 Branches 54 Tags 1.1 GiB
d82a16abd8
Commit Graph

1014 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Jedrzej Hajduczenia
d82a16abd8
[IE CLDNN] Fix fused ops in 1x1 conv fsv16 kernel (#1948) 2020-08-26 20:58:51 +03:00
Vladimir Paramuzov
9618367cff
[IE CLDNN] Fixed memory allocation for case when usm is not available (#1952) 2020-08-26 20:58:21 +03:00
Ilya Churaev
a072c3b7b6
Fixed mean image support for IR v10 (#1926) 
* Fixed mean image support for IR v10

* Fixed comments
 2020-08-26 20:45:54 +03:00
Mikhail Ryzhov
7859aab4b0
Modified samples to support ONNX models (#1680) 
* Modified samples to support ONNX models

- Removed manual loading of IR .bin files (hello_classification, object_detection_sample_ssd)
- Corrected comments
- Added Note in README files

* Modified samples to support ONNX models

- Removed manual loading of IR .bin files (hello_classification, object_detection_sample_ssd)
- Corrected comments
- Added Note in README files
 2020-08-26 18:53:24 +03:00
Artyom Anokhov
6a0f3e5efa
[SCRIPTS] install_NEO_OCL_driver.sh: Added verifying current driver version (#1951) 
* install_NEO_OCL_driver.sh: Added verifying current driver version

* install_NEO_OCL_driver.sh: Updated removal oldest driver on Ubuntu. Updated logic on defining

* install_NEO_OCL_driver.sh: Fixed function name
 2020-08-26 17:25:33 +03:00
Roman Kazantsev
9d5a6cff70
Enable eager execution while TensorFlow 2 model is loaded (#1945) 2020-08-26 17:16:58 +03:00
Anton Pankratv
70839c1663
Fixed threading configuration parameters (#1919) 
* Fixed threading parameters

* Updated documentation

* Fix for macos

* Fix doc
 2020-08-26 16:38:41 +03:00
Vladimir Paramuzov
2e7f4949ca
[IE CLDNN] int8 output support in fsv16 dw conv (#1886) 2020-08-26 16:15:02 +03:00
Kamil Magierski
b77b2e7fbe
Disable CanConvertFromAOTtoSueModel test (#1949) 2020-08-26 16:09:19 +03:00
Anton Pankratv
7c1b87c7aa
Query network should return all supported layers (#1845) 
* Query network should return all supported layers

* Added other plugins. Fixed hetero and multi

* IRv7 test use legacy impl
 2020-08-26 15:33:37 +03:00
Maksim Kutakov
99136f0940
[IE TESTS] Add Reverse Seq single layer test (#1933) 
* [IE TESTS] Add Reverse Seq single layer test

* Added spaces for better readability.
 2020-08-26 15:13:32 +03:00
Sergey Shlyapnikov
fe6cb54bca
[IE CLDNN] Weights reorders optimization (#1542) 
* [IE CLDNN] Fix incorrect (g)_is_os_(z)yx_isv16_osv16 layout name

* [IE CLDNN] Constant blobs copying improvements

* [IE CLDNN] Weights reorders optimization
 2020-08-26 15:12:55 +03:00
Kamil Magierski
4c9fe89487
[GNA] Aligning GNA2 and GNA Unit testing (#952) 
Co-authored-by: kmagiers <kmagiers@intel.com>
 2020-08-26 13:43:43 +03:00
Mateusz Tabaka
a6076a1fd6
Introduce Quantize-Dequantize to FakeQuantize transformation (#1849) 
* Introduce Quantize-Dequantize to FakeQuantize transformation

* Revert changes in DequantizeLinear

* apply code format

* Changes after review:

- description for transformation
- remove NGRAPH_CHECK and move some checks from callback to predicates in pattern
- check if out_low/high are broadcastable for FQ's first input
- fix params to copy_runtime_info

* Add type_matches and type_matches_any predicates

* Use get_single_value

* Changes after review:

- add brief description of transformation
- use get_pattern_value_map instead of get_pattern_map
- change opset1 to opset4
- fix params to copy_runtime_info

* Check result of dynamic_pointer_cast
 2020-08-26 11:51:51 +03:00
Gleb Kazantaev
4673dc9b9c
Fixed Precision Conversion for GenericIE operation type (#1917) 
* Fixed Precision Conversion for GenericIE operation type

* changed logic for setting output type in GenericIE
 2020-08-26 11:31:40 +03:00
Wenzeng Chen
0f524b5523
[IE][VPU]: Fix NCE hang due to input dimensions exceed HW limitation (#1636) 
* Fixed NCE hang due to input dimensions exceed HW limitation
* Added convolution test for big input dimensions

Signed-off-by: wenzengc <wenzeng.chen@intel.com>
 2020-08-26 10:38:58 +03:00
Vladislav Volkov
07ebc2d3bd
Improvements for MKLDNN Input node execution (#1913) 2020-08-26 08:38:29 +03:00
Gleb Kazantaev
12197a4800
Removed WA with adding fake output for VariadicSplit operation (#1938) 2020-08-26 06:28:03 +03:00
Evgenya Stepyreva
c4920ef5a0
[ MO ] Fix for remove_op_node_with_data_node (#1934) 2020-08-25 20:30:41 +03:00
Pavel Rodionov
19570916e9
[GNA] Remove profiler (#1931) 2020-08-25 19:59:01 +03:00
Anton Chetverikov
84c7b048db
SoftPlus operation implementation (#1365) 2020-08-25 19:36:39 +03:00
Ivan Tikhonov
7d1c5877ff
Common specific creator for ops to be converted to ie ops (#1928) 
* Add common specific creator for ops to be converted to ie ops

* fix cpu func tests

* Add macro for specific creators
 2020-08-25 19:23:09 +03:00
Maxim Vafin
b9f698b456
Fix output name collision for incorrect ONNX models (#1930) 2020-08-25 18:43:43 +03:00
Mateusz Tabaka
7b79db8cc8
Adjust tolerance values in ModelImportRunner to the ones onnxruntime … (#1894) 2020-08-25 16:12:55 +02:00
Evgeny Lazarev
aca452def8
Fixed order of transformation to convert the TF OD API SSD models (#1887) 
* Fixed order of transformation to convert the TF OD API SSD models

* Refactored the sub-graph modification for the TF OD API models related to Squeeze/Reshape after SSD heads
 2020-08-25 16:39:34 +03:00
Liubov Batanina
c5b19aa8f9
Added BatchNormInference tests (#1927) 2020-08-25 16:28:05 +03:00
Ilya Lavrenov
03cd918a30
Deprecated migration to Core guide (#1910) 2020-08-25 16:00:09 +03:00
Maksim Doronin
ead37fbebd
[IE][VPU]: Update firmware (#1915) 
* Firmware update contains changed conditions for removing inner dimension with length 1 in Permute kernel
 2020-08-25 11:43:24 +03:00
Maksim Doronin
9000dd8be9
[IE][VPU]: Fix KW issues on Linux (#1916) 
* nullptr dereference fixed in Concat and Reduce stages in Graph Transformer
* missed parameters in asserts in StridedSlice stage was added
 2020-08-25 11:42:23 +03:00
Edward Shogulin
ee96595d1d
[LPT] BLOCKED layout ignoring (#1873) 
* [LPT] BLOCKED layout ignoring

* [LPT] Permute transformation tests

* [LPT] Tests rework
 2020-08-25 10:43:33 +03:00
Alexandra Sidorova
a2f0eef6aa
[CPU] Added H-Swish activation (#1445) 2020-08-25 10:19:06 +03:00
Adam Osewski
393e9295cd
Reference implementation for ScatterUpdate (#1678) 
* Reference implementation for ScatterUpdate and use of it in evaluate.

* Review comments. Clarify comments.

* Update file directory.

* Replace scatter_update reference implementation in ngraph/core/reference/

* Remove template code from ScatterUpdate reference implementation

* Apply review requests

Co-authored-by: mitruska <katarzyna.mitrus@intel.com>
 2020-08-25 06:12:39 +03:00
Edward Shogulin
db2e5c0728
[LPT] GemmTransformation usage fix (#1899) 2020-08-24 20:09:01 +03:00
Tomasz Dołbniak
4013649fe8
v0::Pad operator removal (#1912) 
* v0::Pad operator removal

* Obsolete test removal
 2020-08-24 19:03:01 +03:00
Pavel Rodionov
0ad27cb527
[GNA] Change GNA library version to 1047.1 (#1867) 2020-08-24 18:20:31 +03:00
Roman Kazantsev
76d2a1c351
Disable eager execution affecting TensorFlow 1 concept model handling (#1914) 
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
 2020-08-24 18:01:45 +03:00
Konrad Dobros
f927d71ce4
[IE CLDNN] Fix concat in place with conv interactions (#1714) 
This change fixes concatenation in place optimization where it may
interact with convolution that uses physical padding.
One of such cases is where input to optimized concatenation is also
input to convolution, so it should have padding to enable optimized
implementation.
Previously for all concatenation inputs padding was overriden with only
concatenation axis being padded.
This change fixes this issue by propagating padding across inputs and
output.
 2020-08-24 17:37:46 +03:00
Tomasz Socha
c19cd4765b
Add reshaping of dynamic network in python tests (#1850) 2020-08-24 13:58:38 +02:00
Anastasia Kuporosova
d871a0e05f
[Python API] Deprecation of DataPtr.creator_layer and DataPtr.input_to properties (#1905) 2020-08-24 14:57:55 +03:00
Ivan Tikhonov
cb72684388
Partial revert a commit with reference implementations of PriorBox(Clustered) ops (#1878) 
* partial revert a commit with reference implementations of PriorBox(Clustered), disable references for this ops

* ngraph codestyle

* disable const folding unit tests for PriorBox(Clustered)

* fix arm build

* fix unit test

* Revert "fix unit test"

This reverts commit 1fe59e55d6.
 2020-08-24 14:31:32 +03:00
Evgenya Stepyreva
e6c371ae2e
[ MO ] Reinfer shape sub-graphs once (#1908) 
* [ MO ] Reinfer shape sub-graphs once

* feedback

* feedback
 2020-08-24 14:30:41 +03:00
Liubov Batanina
719797326b
Added test for Pad (#1902) 
* Addded test for Pad

* Add builder
 2020-08-24 14:16:29 +03:00
Ivan Tikhonov
c74643f6b6
Fix Assign, Matmul ops for correct work with keep_shape_ops feature for Kaldi models (#1885) 
* fix assign and matmul ops

* fix unit test

* static shapes by default for kaldi
 2020-08-24 14:11:01 +03:00
Vladimir Paramuzov
2acf3f27e1
[IE CLDNN] Enabled tensor pitches in topk kernel for 0 input (#1898) 2020-08-24 13:03:39 +03:00
Lukasz Debski
0a5dfc62b2
[IE CLDNN] Forcing bfzyx format in MVN layer (#1903) 2020-08-24 13:03:21 +03:00
Mateusz Tabaka
d7db4974f3
Reduce number of ops needed to create InstanceNorm (#1896) 
* Reduce number of ops needed to create InstanceNorm

InstanceNorm in onnx importer creates the same subgraph for Mean twice - once for Variance and once for actual Mean.
This change makes InstanceNorm to use single Mean which is shared by numerator and Variance.

Also enables IE_CPU.onnx_model_instance_normalization test case

* Revert changes to .gitignore
 2020-08-24 12:26:15 +03:00
Vladimir Paramuzov
c8aa3a7f2b
[IE CLDNN] FP16 support for some quantized kernels (#1855) 2020-08-24 08:47:45 +03:00
Tomasz Dołbniak
d4567d5ab5
Linux clang compilation error fix (#1891) 2020-08-21 17:22:54 +03:00
Evgenya Stepyreva
80e09dfa3b
[ MO ] Change layout of Shape sub-graphs once (#1875) 
* [ MO ] Change layout of Shape sub-graphs once
 2020-08-21 16:27:02 +03:00
Tomasz Dołbniak
067d29a3f0
Accept empty initializers from ONNX models (#1776) 
* Accept empty initializers from ONNX models

* Create dummy constants instead of null nodes

* Code formatting

* Test the empty initializers handling in onnx importer
 2020-08-21 15:15:27 +03:00