NeuralNetworkInference plugin:
Setting EnablePattern in CPU and disable for DML. Remove runningtime warning:
[W:onnxruntime:, inference_session.cc:489 onnxruntime::InferenceSession::RegisterExecutionProvider] Having memory pattern enabled is not supported while using the DML Execution Provider. So disabling it for this session since it uses the DML Execution Provider.
#rb Gines.Hidalgo, Nebojsa.Dragosavac
[CODEREVIEW] Gines.Hidalgo, Nebojsa.Dragosavac
#preflight 61b1210de01bc44973fb909b
#ROBOMERGE-AUTHOR: francisco.vicente
#ROBOMERGE-SOURCE: CL 18413072 in //UE5/Release-5.0/... via CL 18413092
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v897-18405271)
[CL 18413108 by francisco vicente in ue5-release-engine-test branch]
- Removed OutputDeviceType almost completely from FBackEndUEAndORT because it is not used anywhere.
- ENeuralTensorType TensorType is now being properly set to either ENeuralTensorType::Input or ENeuralTensorType::Output (never to Generic).
- Run() now uses ENeuralDeviceType InputDeviceType to know whether to do the initial GPU sharing, no need to use ENeuralTensorType's workaround anymore.
[at]Neb, this one properly fixes the initial workaround that was made to make it work, and stops wrongly using ENeuralTensorType as a GPU vs CPU field and lets ENeuralDeviceType take care of that (which was created for that purpose anyway). Please, let me know if you find any issues with it!
#rb nebojsa.dragosavac
[CODEREVIEW] nebojsa.dragosavac
#rnx
#ROBOMERGE-AUTHOR: gines.hidalgo
#ROBOMERGE-SOURCE: CL 18403192 in //UE5/Release-5.0/... via CL 18403193
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v896-18170469)
[CL 18403194 by gines hidalgo in ue5-release-engine-test branch]
- Several functions removed or their arguments re-ordered to be more unified and follow the very same style.
- More functions with doxygen doc.
- Updated all classes/functions that used those modified/removed functions to keep working.
- GetPooledBuffer made constant.
#rb Francisco.Vicente
[FYI] Francisco.Vicente
#rnx
#ROBOMERGE-AUTHOR: gines.hidalgo
#ROBOMERGE-SOURCE: CL 18402860 in //UE5/Release-5.0/... via CL 18402862
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v896-18170469)
[CL 18402864 by gines hidalgo in ue5-release-engine-test branch]
NeuralNetworkInference plugin:
Fix note: see previous definition of 'TEXT' in dml_provider_factory.cc in ORT 11/30
#rb Gines.Hidalgo
[CODEREVIEW] Gines.Hidalgo
#ROBOMERGE-AUTHOR: francisco.vicente
#ROBOMERGE-SOURCE: CL 18399949 in //UE5/Release-5.0/... via CL 18399982
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v896-18170469)
[CL 18399995 by francisco vicente in ue5-release-engine-test branch]
[at]Neb, please, let me know if there are any issues with this function, I could not find any. GetName is now returning a reference, so it's no longer going out of scope as you faced in the past when you created GetNameData as fix, but let me know if I am wrong!
#rb Francisco.Vicente, nebojsa.dragosavac
[CODEREVIEW] Francisco.Vicente, nebojsa.dragosavac
#rnx
#ROBOMERGE-AUTHOR: gines.hidalgo
#ROBOMERGE-SOURCE: CL 18397076 in //UE5/Release-5.0/... via CL 18397080
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v896-18170469)
[CL 18397097 by gines hidalgo in ue5-release-engine-test branch]
This allows to list quickly in code what shader class are left to refactor to RDG
#rb sebastien.hillaire
#jira none
#preflight 61ae5ab8fc6bcc8e19712326
#ROBOMERGE-AUTHOR: guillaume.abadie
#ROBOMERGE-SOURCE: CL 18386269 in //UE5/Release-5.0/... via CL 18386288
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v896-18170469)
[CL 18386315 by guillaume abadie in ue5-release-engine-test branch]
- Moved NNIProfiler and NNIQA into "Private" (as the client do not need them).
#rb Francisco.Vicente
[CODEREVIEW] Francisco.Vicente
#rnx
#ROBOMERGE-AUTHOR: gines.hidalgo
#ROBOMERGE-SOURCE: CL 18346967 in //UE5/Release-5.0/... via CL 18346975
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v895-18170469)
[CL 18346989 by gines hidalgo in ue5-release-engine-test branch]
NNI/Public/NeuraTensor.h reduced to up to 150-characters per line
#rb Francisco.Vicente
[CODEREVIEW] Francisco.Vicente
#rnx
#ROBOMERGE-AUTHOR: gines.hidalgo
#ROBOMERGE-SOURCE: CL 18345590 in //UE5/Release-5.0/... via CL 18345600
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v895-18170469)
[CL 18345611 by gines hidalgo in ue5-release-engine-test branch]
- NNI/Public/NeuralNetwork.h cleaned a bit
- All Enum classes moved to enumclasses.h (not renamed yet).
#rb Francisco.Vicente
[CODEREVIEW] Francisco.Vicente
#rnx
#ROBOMERGE-AUTHOR: gines.hidalgo
#ROBOMERGE-SOURCE: CL 18333818 in //UE5/Release-5.0/... via CL 18333827
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v895-18170469)
[CL 18333841 by gines hidalgo in ue5-release-engine-test branch]
