064a18eb09
[POT] Change python version to 3.8+ ( #20071 )
...
* Change python version to 3.8+
* Fix tests
2023-09-27 18:01:34 +04:00
a4b0fe51af
[POT] Update for CI ( #19067 )
...
* Update references
2023-09-01 18:12:24 +02:00
5fca707ebd
[POT] Update CPU_SPR transformer quantization ( #15241 )
...
* [POT] Update CPU_SPR transformer quantization
* Add comments
2023-01-24 15:56:22 +04:00
39ff705a3e
[POT] Update CI branch ( #14916 )
...
Co-authored-by: Alexander Kozlov <alexander.kozlov@intel.com>
2023-01-03 17:26:36 +00:00
7ceb8bfce0
[POT] Update test references ( #14318 )
...
* Update references
* Update GNA references
* Rebase branch
2022-12-07 14:05:03 +04:00
bb00a9e664
[POT] Enable quantization inside subgraphs for CPU ( #12776 )
...
* Enable LSTMCell quantization
* Refactored unify function
* reverse dict
* Fix docs
* Add LSTM and GRU sequences
* Change scales to unified_scales in hw configs
* Fix inplace statistics
* Shift axis for subgraphs
* Changed HW config
* Update HW configs for tests
* Not quantize GRU ops with linear_before_reset: true
* Enable tests
* Support models with scalar input
* Fix tests
* return back submodules
* Fix tests
* Update GPU HW config
* Fix comments
* Fix axis
2022-11-12 07:54:29 +04:00
8e9a6d3457
[POT] BC update for transformers ( #13869 )
...
* Update node_utils
* HBC update
* Updated HBC & added test
* Apply comments
2022-11-11 11:33:11 +04:00
03dbd8c4e7
[POT] Update node names in ONNX models ( #13872 )
2022-11-05 21:00:19 +01:00
02d30c8aaa
[POT] Unify MatMul scales ( #13717 )
...
* Support FQ scales unification for MatMul
* Add test
* Update test_scales.py
2022-11-05 09:33:54 +04:00
8789fcda04
[POT] Propagate FQ through Gather ( #13583 )
...
* Add Gather to QUANTIZE_AGNOSTIC_OPERATIONS
* Add test
* Update HW configs path in tests
* Fix tests
* Update CI references
2022-11-04 12:21:20 +04:00
ee93ddc453
Update references ( #13637 )
2022-10-27 10:55:54 +00:00
968bd6c96b
Fix bug with unknown node type ( #13406 )
2022-10-11 19:05:54 +04:00
9f2564670f
[POT] Support SoftSign quantization for GNA ( #13140 )
...
* Add SoftSign to GNA HW config
* Add test
* Update gna3 HW config
2022-10-02 10:43:51 +00:00
f68b6e27b0
Update references ( #13245 )
2022-09-30 18:10:01 +04:00
4df8ed881a
[POT] Unroll GRUCell for GNA ( #12930 )
...
* [POT] Unroll GRUCell for GNA
* Add test
2022-09-23 21:57:10 +03:00
572d00eaa2
[POT][GNA] add mixed-precision support for Accuracy Aware Quantization ( #12470 )
...
* [POT] add GNA3 mode
* [POT][GNA3] add mixed-precision support for Accuracy Aware Quantization
* [POT] add GNA AAQ sanity test with synthetic data
* [POT][GNA3] add GNA3.5 mode and minimize AAQ option
* [POT] fix pot/ci fails on config test
2022-09-20 08:47:33 +09:00
d0d88fa1ec
[POT] Skip per channel quantizers allining before concats ( #12483 )
...
* Skip per channel quantizers allining before concats
* Test coverage
2022-08-29 13:55:37 +04:00
edaa7f8e7e
[POT] Precommit reference update ( #12304 )
...
* WIP graph tests fixing
* Fix collectiors graph tests
* remove debug code
* Fix rebase
* eps update for scales tests
* Outputs for some reference models was changed
* Sanity reference metrics update for VNNI CI hosts
* Unused hyperopt dependency which broke python3.6 support is commented
* Minor comments fixes
2022-08-05 17:50:45 +04:00
b371e74735
[hotfix] pytest error of act_act example ( #12093 )
...
* [hotfix] pytest error of act_act example
* remove needless import
2022-07-11 12:56:32 +09:00
d5e8d1d968
Nm/outputs quantization scheme ( #12035 )
...
* [POT] outputs quantization scheme
* [POT] remove needless blank line
* add a range_estimator config for outputs
* Add fq_config_priority
2022-07-07 16:08:55 +09:00
99f0093615
[POT] Fixed bug with out port for nodes with multiple outputs ( #10932 )
...
* Fixed bug with out port
* Add test
* Fix test
* Change test model
2022-03-23 11:17:06 +03:00
e103af056e
[POT] References & golds update ( #10937 )
...
* Update references
* Update golds & add stats dumping
* Statistics_data upd
* Enable densenet in nightly
* Pylint fixes
* Update try-except WA
* Update simplified gold
2022-03-15 16:55:48 +03:00
be6a3c34f1
[POT] Throw exception for IRv10 ( #10345 )
...
* [POT] Throw exception for IRv10
* Update reference models
* Updated AC framework name from dldt to openvino
2022-02-17 10:54:08 +03:00
0c855ee8b2
[POT] Renaming NXModel ( #10168 )
...
* NXModel -> CompressedModel renaming
* Update references & remove Dicts
* Pylint fixes
2022-02-08 14:07:12 +03:00
aa518282b0
updat gna configs ( #9920 )
2022-01-27 08:50:15 +03:00
ed634bd18e
[POT] Tests & References update ( #9694 )
...
* Enable disabled tests
* Update references
* Update references
* Update passes.py for FP16 graph
* Update precision for nightly mobilenet
* Update references
* Update refs and disable model
* Update nightly refs
* Update nightly
* Revert mbv2 sanity gold
2022-01-20 16:23:00 +03:00
8fe5484645
[Python API] Remove offline transformations from old python api ( #9121 )
...
* [Python API] Remove offline transformations from old python api
* try to fix import error
* try to fix pylint
* try to fix pylint2
* Use new api in graph_utils
* Fix pylint
* Try to fix pylint
* Use serialize from pass manager
* try to skip tests
* try to use new ir
Co-authored-by: AlexeyLebedev1 <alexey.lebedev@intel.com>
2022-01-10 17:32:55 +03:00
950fc508ad
[POT] Update tests references ( #9187 )
...
* Update test utils with the new MO
* Update references
* Refs update
* Disabled few tests & update refs
* Change model data_type
* Added TODOs
2021-12-14 12:15:49 +03:00
2d0ae6028a
[POT] Added quantization inside subgraphs ( #8530 )
...
* [POT] Support subgraphs
* Add support for inplace statistics
* Fixed cascade models
* Fixed bug
* Fixed import
* Fix name
* Add test
* Fix comments
* Replace test model
* Replace test model
* Add recursively flag
* Fix pylint
* Updated AA node names
* Fix adding results to node
* Disable subgraph quantization
* Disable recursion search in graph
* Skip test with subgraphs quantization
2021-11-26 12:42:33 +03:00
516d510045
[POT] Added typecast while nodes updating & creation ( #8621 )
...
* Update type casting while update
* Fix Pylint
* Update precision setting for FQ values
* Update FQ values precision setting
* Fix scales dump
* Update reference
* Another reference update
2021-11-23 17:13:04 +03:00
29a3f56003
[POT] Add pattern se blocks ( #8425 )
...
* fix(patterns): add pattern se_block
* update reference
2021-11-15 12:53:07 +03:00
8cb7824c21
[POT] Update for the Results quantization & tests ( #8324 )
...
* Update rule for multi-results quantization
(cherry picked from commit d94fe7d75816fe385117
2021-11-08 16:28:54 +03:00
ce9a968030
[POT] Update tests with new data ( #8209 )
...
* Update POT tests with new data
* Revert package changes
* Remove mobinenet-ssd from POT tests
* Update benchmark.py IE Python API usage
2021-10-27 12:40:19 +03:00
bbeec714aa
Moved Post-training Optimization Tool to open-source ( #7940 )
...
* Moved POT to opensource
* Added OMZ as a submodule
* Exclude OMZ from ShellCheck
2021-10-15 16:35:35 +03:00
Nikita Malinin