WebML WG Teleconference – 14 October 2021 - 14:00-15:00 UTC+0

Latest version: https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md


WebML WG Teleconference – 14 October 2021 - 14:00-15:00 UTC+0
See the timezone table ...
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-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/2021/10/14-webmachinelearning-minutes.html


________________________________
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#agenda>Agenda
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#%E2%84%B9%EF%B8%8F-webnn-api-recent-new-feature-requests>ℹ️ WebNN API recent new feature requests

Discuss and triage new feature requests:

  *   Autograd / JAX proposal for future work https://github.com/webmachinelearning/webnn/issues/218


<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#%E2%84%B9%EF%B8%8F-tpac-sessions-published>ℹ️ TPAC sessions published

Review our WebML WG Virtual Meeting sessions: https://github.com/webmachinelearning/meetings/issues/18


Please do not hesitate to self-nominate yourself as a contributor for any sessions of interest and share relevant resources for background reading.

Summary:

26 October 2021 14:00-15:00 UTC+0

  *   Rationale/criteria for adding new ops to the WebNN API
  *   Versioning and web compatibility
  *   Privacy discussion

27 October 2021 14:00-15:00 UTC+0

  *   ML JS framework performance, focus areas for WebNN
  *   Integrating an open-source cross-platform implementation of the Web Neural Network API into a web engine
  *   How should WebNN API deal with fusion

28 October 2021 14:00-15:00 UTC+0

  *   Conformance testing of WebNN API
  *   Ethical issues in using Machine Learning on the Web

<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#%E2%84%B9%EF%B8%8F-appendix-help-wanted>ℹ️ Appendix: Help wanted

These issues on our issue tracker are welcoming comments and proposed solutions. We discuss these issues on our bi-weekly meetings time permitting. To bump the priority, please chime in on the issues.

<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#proposed-new-features>Proposed new features

Discuss and triage new feature requests:

  *   BatchNormalization should be an optional operation https://github.com/webmachinelearning/webnn/issues/187

  *   BatchToSpaceND and SpaceToBatchND ops https://github.com/webmachinelearning/webnn/issues/189

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

  *   argMax op https://github.com/webmachinelearning/webnn/issues/184

  *   Dilated pooling https://github.com/webmachinelearning/webnn/issues/180

  *   int8 quantized https://github.com/webmachinelearning/webnn/issues/128

  *   Dynamic shape inference https://github.com/webmachinelearning/webnn/issues/124

  *   QuantizeLinear and DequantizeLinear https://github.com/webmachinelearning/webnn/issues/93


<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#to-be-clarified>To be clarified

  *   OperandType of gemm / matmul return https://github.com/webmachinelearning/webnn/issues/84

  *   Handling unsupported OperandType https://github.com/webmachinelearning/webnn/issues/36

  *   OperandDescriptor's dimension should match ECMAScript's TypedArray dimension https://github.com/webmachinelearning/webnn/issues/29


<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#ready-to-close>Ready to close?

  *   Specify the ModelBuilder.createModel and other ModelBuilder members https://github.com/webmachinelearning/webnn/issues/107


<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#investigation>Investigation

  *   Buffer sharing between GPU and ML accelerator https://github.com/webmachinelearning/webnn/issues/33


<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#explainer-updates>Explainer updates

  *   Workflow from authoring tools to webnn https://github.com/webmachinelearning/webnn/issues/74


<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#new-use-cases>New use cases

  *   Neural network deployment https://github.com/webmachinelearning/webnn/pull/207


<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#webidl-conventions>WebIDL conventions

  *   Add method steps to operations https://github.com/webmachinelearning/webnn/issues/210

  *   Define algorithms for dictionaries with lists as default values https://github.com/webmachinelearning/webnn/issues/211


<https://github.com/webmachinelearning/meetings/blob/main/telcons/2021-10-14-agenda.md#deferred-to-v2>Deferred to v2

  *   Support output tensor in native memory format https://github.com/webmachinelearning/webnn/issues/173