hpc-lab

AI-HPC tasks studies.

This year, we focus on following tasks to seek out best-practice over Taiwania-2 in TWCC.

HPC	Vision	NLP	STT/TTS	other Tasks
Climate segmentation	ResNet-50 v1.5	RNN-T	Kaldi	DLRM
Cosmology parameter prediction	3D U-Net	BERT-large	Tacotron 2 and WaveGlow	Mini Go
Quantum molecular modeling	SSD	ELECTRA		RAPIDS
	Mask R-CNN
	YOLO v4
	MobileNetV3
	Efficient Net
	Inception v3
	UNet Medical Image Segmentation

Tasks will follow MLCommons and NVIDIA Deep Learning, including:

• MLPerf Train HPC track (3 tasks)
• MLPerf Train track (8 tasks)
• NVIDIA Deep Learning (9 selected tasks)

All results and percedures will be documented accordingly.