* ChromeOS enabling changes in DLDT/OPENVINO
Changes for new activations
Enable Debug messages
Add exp activation files
Enable FP32 functions for newly added activations
Enable activation names in debug files
Modify Scale Factor calcuation
* Add support for SetState in GNAMemory
* Use rounding when changing the scale + switch to Elevoc's identity
* Merge branch 'private/kmagiers/GNAPlugin_Memory_layer_as_output' into 'master'
Input memory layer support in addOutput();
See merge request inference-engine/dldt!7016
* Porting Dldt to ChromeOS
Signed-off-by: Anisha Kulkarni <anisha.dattatraya.kulkarni@intel.com>
* Optimizations for builder & GNAPlugin
- Optimize graph building addLayer
- Cache Activation functions pwl
- Use AVX/SSE intrinsics for Quantization
- Add config to add Identity layer Scale Factor
- Port Permute related changes, Permute
operation is needed for Layer Normalization
batching.
- Additionally, enable Negetive Half Log
Activation Layer
* Work Around for DivByN and permute
- Scale Factors need further tuning before
the DivByN operation in GNA Layer Norm
- Work around for Permute
* BACKPORT:Enable CoreThreadingTestsWithIterations tests for GNA
Co-authored-by: Raviraj P Sitaram <raviraj.p.sitaram@intel.com>
Co-authored-by: Denis Orlov <denis.orlov@intel.com>
Co-authored-by: Smirnov, Eugene <eugene.smirnov@intel.com>
Co-authored-by: Anisha Kulkarni <anisha.dattatraya.kulkarni@intel.corp-partner.google.com>
83964338b0