WebML CG Teleconference – 31 October 2019 - 15:00-16:00 UTC+0

https://github.com/webmachinelearning/meetings/blob/master/telcons/2019-10-31-agenda.md

Note: Exceptional schedule for this call on the following locations:

08:00 PDT San Francisco (U.S.A. - California) 
11:00 EDT Boston (U.S.A. - Massachusetts)		 
23:00 CST Shanghai (China)			        
00:00 JST Tokyo (Japan)				        

(Between October 27 and November 3, the time difference between US and Europe will be 1 hour less than it currently is. China, Japan and Russia do not observe DST, which means the time difference between these areas and US/Europe will be one hour more than it currently is.)

Logistics

* Chair: Anssi
* Scribe: ?
* IRC: irc://irc.w3.org:6667/#webmachinelearning
* IRC web client: http://irc.w3.org/?channels=#webmachinelearning
* Call-in details: https://lists.w3.org/Archives/Member/internal-webmachinelearning/2019Apr/0000.html
* Minutes: https://www.w3.org/2019/10/31-webmachinelearning-minutes.html

Agenda

* WebNN interop investigation
- Review POC requirements for buffer sharing between dedicated ML hardware and GPU (Ningxin, Paul)

* Op compatibility
- Review concrete spec proposals for conv2d #28, matMul #27
  https://github.com/webmachinelearning/webnn/issues/28
  https://github.com/webmachinelearning/webnn/issues/27

* Explainer
- Discuss how to move explainer.md forward, priorities
  https://github.com/webmachinelearning/webnn/blob/master/explainer.md

* On-device training (exploratory topic)
- Discuss key use cases, feasibility, platform support (Core ML 3, ONNX 1.7 ...)
- Note: currently out of scope, a future opportunity subject to group agreement

Thanks,

-Anssi (WebML CG chair)

Received on Thursday, 24 October 2019 07:04:02 UTC