Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
43 lines (40 sloc) 3.03 KB
amazon_s3_presentation_url amazon_s3_video_url author categories comments date image layout session_id session_track slideshare_presentation_url speakers title youtube_video_url tag
2018-09-16 09:00:00+00:00
featured file_name path
Machine Learning/AI, AI and Neural Networks on Arm Summit
biography company job-title name speaker-image
"Luba Tang received his M.S. degree in computer science in only one year from the National Tsing-Hua University, Taiwan. He has been a Ph.D. student in computer science department of National Tsing-Hua University, Taiwan since 2007. At the same time, he has been working in the compiler groups in Marvell, Inc. and MediaTek, Inc. since 2010. His research interests include both electronic system level (ESL) design and compilers. He had focused on iterative compiler, ahead-of-time compiler, link-time optimization, electronic system level simulation, and electronic system level design. His most recent work focus is on dynamic linking and link-time optimization. He was the chief programmer of Starfish DSP simulator, the original writer of Marvell iterative compiler, and the software architect of MCLinker and ONNC. He is also the founder and CEO of Skymizer Inc."
Luba Tang
YVR18-333: ONNC (Open Neural Network Compiler) for Arm Cortex-M

The Open Neural Network Compiler (ONNC) project aims to provide a compiler to connect Open Neural Network Exchange Format (ONNX) to every deep learning accelerator (DLA). ONNX is a standard format for representing deep learning models that enables models to be correctly transferred between frameworks, like caffe, CNTK, mxnet, pytorch and TensorFlow. ONNX guarantees interoperability between frameworks, however, the industry still needs a backer to guarantee executability between DLAs - to ensure every DLA can execute ONNX models correctly.

ONNC is such a backer for DLA vendors. It is a kind of cross compiler that transforms ONNX models into binary machine code of DLAs. Every DLA has its own unique and delicate design in its memory for fast data movement. A compiler must provide sufficient flexibility to handle with the wide range of varieties. ONNC leverages the IR design of ONNX and provides rich and effective algorithms to eliminate overhead of data movement. And the best is that DLA vendors can easily reuse these algorithms by just describing its own unique physical cost model. Skymizer hopes DLA vendors can be free from re-inventing these intricated optimization algorithms.

In this talk, we not only introduce ONNC framework, we will dive into ONNC internals. We will explain our plan to support uTensor backend for Arm Cortex-M and discuss some technical issues.

You can’t perform that action at this time.