WebML CG Teleconference – 27 May 2021 - 14:00-15:00 UTC+0

Latest: https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-05-27-agenda.md

WebML CG Teleconference – 27 May 2021 - 14:00-15:00 UTC+0
San Francisco (U.S.A. - California)     Thu, 27 May 2021        07:00   UTC-7 hours
Boston (U.S.A. - Massachusetts) Thu, 27 May 2021        10:00   UTC-4 hours
London (United Kingdom - England)       Thu, 27 May 2021        15:00   UTC+1 hours
Berlin (Germany)        Thu, 27 May 2021        16:00   UTC+2 hours
Helsinki (Finland)      Thu, 27 May 2021        17:00   UTC+3 hours
Shanghai (China)        Thu, 27 May 2021        22:00   UTC+8 hours
Tokyo (Japan)   Thu, 27 May 2021        23:00   UTC+9 hours
Corresponding UTC (GMT) Thu, 27 May 2021        14:00 UTC

Other locations: https://www.timeanddate.com/worldclock/fixedtime.html?iso=20210527T14


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-05-27-agenda.md#logistics>Logistics

  *   Chair: Anssi
  *   Scribe: ?
  *   IRC: irc://irc.w3.org:6667/#webmachinelearning
  *   IRC web client: https://irc.w3.org/?channels=#webmachinelearning

  *   Call-in details: https://lists.w3.org/Archives/Member/internal-webmachinelearning/2020Apr/0000.html

  *   Minutes: https://www.w3.org/2021/05/27-webmachinelearning-minutes.html


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-05-27-agenda.md#agenda>Agenda
<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-05-27-agenda.md#security-and-privacy>Security and Privacy

Review and discuss initial Security and Privacy Considerations:

  *   Issue: https://github.com/webmachinelearning/webnn/issues/122

  *   PR: https://github.com/webmachinelearning/webnn/pull/170

  *   Related issues:
     *   [privacy-tracker] Self-Review Questionnaire: https://github.com/webmachinelearning/webnn/issues/119

     *   [privacy-tracker] Fingerprinting via matmul: https://github.com/webmachinelearning/webnn/issues/85


Discuss related WebGPU/GL Security and Privacy Considerations, applicability in this context:

  *   WebGPU: https://gpuweb.github.io/gpuweb/#security

  *   WebGL: https://www.khronos.org/registry/webgl/specs/latest/1.0/#4


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-05-27-agenda.md#operation-specific-apis-proposal>Operation-specific APIs proposal

Continue progress remaining features that satisfy the requirements of the operation-specific APIs proposal:

https://github.com/webmachinelearning/proposals/issues/2


Discuss design considerations and review proposed solutions to enable both efficient graph execution and imperative eager execution with a cohesive WebNN API. Related topics:

  *   Support download data asynchronously
     *   PR: WIP https://github.com/webmachinelearning/webnn/pull/166

  *   Layout format conversion
     *   Issue: WIP
  *   Clarify at which point weights are used in the compilation
     *   Issue: https://github.com/webmachinelearning/webnn/issues/157

  *   AI accelerator device selection
     *   Issue: https://github.com/webmachinelearning/webnn/issues/169


<https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-05-27-agenda.md#tag-review>TAG review

  *   [tag-tracker] Define a common term for logical tensor changes?
     *   Issue: https://github.com/webmachinelearning/webnn/issues/150

     *   PR: WIP
  *   [tag-tracker] Ergonomics of the JS examples
     *   Issue: https://github.com/webmachinelearning/webnn/issues/139

     *   Resolution: WIP - pending concrete TAG feedback to inform the resolution
  *   [tag-tracker] String enum for activations
     *   Issue: https://github.com/webmachinelearning/webnn/issues/138

     *   Proposed resolution: WIP - review informative note to be incorporated into the spec to explain current design principles around activations