Collective Knowledge workflows for MLPerf

Linux/MacOS:

MLPerf Inference v0.5

Task	Model	Owner	Contributors	Implementation	Environment (need more testing)	CK workflow
Object Classification	MobileNet	dividiti	dividiti	TFLite	Linux, Android (Windows, Mac)	yes
			dividiti	TF (C++)	Linux, Android (Windows, Mac)	yes
			dividiti	TF (Python)	Linux (Windows, Mac)	yes
			dividiti, NVIDIA, Microsoft	ONNX	Linux, Mac (Windows)	yes
			dividiti, Arm	ArmNN	Linux (Android)	yes
	ResNet	Microsoft	Microsoft	ONNX	Linux (?)	no
			Microsoft	TF (Python)	Linux (?)	no
			dividiti	TFLite	Linux, Android (Windows, Mac)	yes
			dividiti	TF (C++)	Linux, Android (Windows, Mac)	yes
			dividiti	ONNX	Linux, Mac (Windows)	yes
			dividiti, Arm	ArmNN	Linux (Android)	yes
Object Detection	SSD-MobileNet	dividiti	dividiti	TFLite	Linux (Android, Windows, Mac)	yes
			dividiti, Habana	TF (Python)	Linux (?)	yes
			Facebook	PyTorch	Linux (?)	no
	SSD-ResNet	Habana	Habana, NVIDIA	PyTorch	Linux (?)	no
			GM, Google, dividiti	TF	(?)	coming soon!
Machine Translation	GNMT	Intel	Intel	TF (Python)	Linux (?)	no
			Intel	PyTorch	Linux (?)	no

NB: We generally acknowledge Google for making many pretrained TF models available (e.g. for Object Detection).

MLPerf Training v0.6

TODO