WebML CG Teleconference – 21 January 2021 - 15:00-16:00 UTC+0

https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md


WebML CG Teleconference – 21 January 2021 - 15:00-16:00 UTC+0
San Francisco (U.S.A. - California)     Thu, 21 January 2021    07:00   UTC-8 hours
Boston (U.S.A. - Massachusetts) Thu, 21 January 2021    10:00   UTC-5 hours
London (United Kingdom - England)       Thu, 21 January 2021    15:00   UTC+0 hours
Berlin (Germany)        Thu, 21 January 2021    16:00   UTC+1 hours
Helsinki (Finland)      Thu, 21 January 2021    17:00   UTC+2 hours
Shanghai (China)        Thu, 21 January 2021    23:00   UTC+8 hours
Tokyo (Japan)   Fri, 8 January 2021     00:00   UTC+9 hours
Corresponding UTC (GMT) Thu, 21 January 2021    15:00 UTC

Other locations: https://www.timeanddate.com/worldclock/fixedtime.html?iso=20210121T15


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md#logistics>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/2020Apr/0000.html (login with your W3C account<https://www.w3.org/Help/Account/>) - Zoom app install<https://zoom.us/download> recommended
  *   Minutes: https://www.w3.org/2021/01/21-webmachinelearning-minutes.html


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md#agenda>Agenda
<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md#webnn-native>webnn-native

Present the proposed standalone native implementation of the WebNN API spec called webnn-native that complements the JS implementation webnn-polyfill. Both the implementations help inform the API spec development, validate implementability atop native APIs (initial targets OpenVINO and DirectML, extensible with other backends), and provide a performance benchmark. (Ningxin)

Discuss the proposal to adopt webnn-native as a CG deliverable similarly to webnn-polyfill. No CG Charter changes needed, is considered "Other Software" per the existing charter text.

<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md#webnn-conv2d-layout-parameter-tensorflow-incompatibility>WebNN conv2d layout parameter TensorFlow incompatibility

Discuss and gather further feedback on TF, TF Lite, and TF.js preference for input and filter layout approach in WebNN API:

https://github.com/webmachinelearning/webnn/issues/125


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md#numerical-precision-in-conformance-testing>Numerical precision in conformance testing

Discuss and solicit insights on difference between TF.js WebGL and CPU/Wasm backends:

https://github.com/webmachinelearning/webnn-polyfill/pull/32


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md#nsnet2-sample-and-tfjs-memory-leak>NSNet2 sample and TF.js memory leak

Check TF.js upstream blocker status and discuss any other opens blocking PR:

https://github.com/webmachinelearning/webnn-samples/pull/22


https://github.com/tensorflow/tfjs/issues/4378


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md#proposals-for-future-work>Proposals for future work

Continue discuss proposals submitted for consideration for future work:

  *   Data processing: https://github.com/webmachinelearning/proposals/issues/1


  *   Operation-specific APIs: https://github.com/webmachinelearning/proposals/issues/2


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-01-21-agenda.md#security-and-privacy-considerations>Security and Privacy considerations

Last Call to review the proposed responses to the Self-Review Questionnaire for Security and Privacy:

https://github.com/webmachinelearning/webnn/issues/119