WebML WG Teleconference – 24 April 2025 - 15:00-16:00 UTC

Latest version: https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md


WebML WG Teleconference – 24 April 2025 - 15:00-16:00 UTC
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#webml-wg-teleconference--24-april-2025---1500-1600-utc>
❗ We're starting 08:00 Pacific Daylight Time, see also other locations<https://www.timeanddate.com/worldclock/fixedtime.html?iso=20250424T15>. ❗
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#-were-starting-0800-pacific-daylight-time-see-also-other-locations->
See the timezone table ...
San Francisco   Thu, 24 April 2025      08:00
Boston  Thu, 24 April 2025      11:00
London  Thu, 24 April 2025      16:00
Berlin  Thu, 24 April 2025      17:00
Helsinki        Thu, 24 April 2025      18:00
Shanghai        Thu, 24 April 2025      23:00
Tokyo   Fri, 25 April 2025      00:00
UTC     Thu, 24 April 2025      15:00 UTC

Other locations: https://www.timeanddate.com/worldclock/fixedtime.html?iso=20250424T15

Logistics
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#logistics>

  *   Chair: Anssi
  *   Scribe: Anssi
  *   IRC: irc://irc.w3.org:6667/#webmachinelearning

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

  *   Zoom joining instructions: https://lists.w3.org/Archives/Member/internal-webmachinelearning/2023Jun/0000.html

  *   Minutes: https://www.w3.org/2025/04/24-webmachinelearning-minutes.html


Agenda
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#agenda>
🧪 Incubations
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#-incubations>

Discuss recent WebML Community Group development to keep the Working Group abreast of incubation progress.

🤝 WG-CG collaboration
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#-wg-cg-collaboration>

Solicit input for a proposal to combine WG and CG meetings to encourage broader participation.

🥡 BlinkOn 20 takeaways
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#-blinkon-20-takeaways>

Share relevant BlinkOn takeaways. Map feedback & tidbits to solutions and issues.

  *   Compute Abstraction for AI: Wasm, WebGPU, and WebNN<https://www.youtube.com/watch?v=IgIdayJH4_o>
     *   Expose available hardware → device selection<https://github.com/webmachinelearning/webnn/labels/device%20selection>
     *   Ahead-of-time compilation → MLGraph caching<https://github.com/webmachinelearning/webnn/issues/807>
     *   NPU differences & model portability → op support limits<https://www.w3.org/TR/webnn/#api-mlcontext-opsupportlimits> future work?
     *   Hybrid execution → best-effort buffer-sharing with tensors<https://github.com/webmachinelearning/webnn/blob/main/mltensor-explainer.md> and constants<https://github.com/webmachinelearning/webnn/pull/830>
  *   Expandable Built in AI: Opening the Vision with Shared AI Models<https://www.youtube.com/watch?v=3zOYVlBKOHA>
     *   Common model format & ops → (paused) Model Loader API<https://github.com/webmachinelearning/model-loader>, core operator set<https://github.com/webmachinelearning/webnn/issues/573>
     *   Model sharing → built-in models<https://github.com/webmachinelearning/webnn/issues/1> exposed to low-level compute APIs?
     *   Built-in model transparency → Model Cards<https://arxiv.org/abs/1810.03993> integration?

🏷️ Operator specific<https://github.com/webmachinelearning/webnn/labels/operator%20specific> issues
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#%EF%B8%8F-operator-specific-issues>

Review and discuss operator specific issues that reduce code complexity and improve maintainability, e.g.:

  *   pad
     *   webmachinelearning/webnn#377<https://github.com/webmachinelearning/webnn/issues/377>
     *   webmachinelearning/webnn#739<https://github.com/webmachinelearning/webnn/issues/739>

 Note

We will discuss other operator specific<https://github.com/webmachinelearning/webnn/labels/operator%20specific> matching the criteria based on participants' suggestions, time allowing.

ℹ️ Tensors for graph constants
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#%E2%84%B9%EF%B8%8F-tensors-for-graph-constants>

Discuss PR review feedback and proposed API design alternatives on how to best represent constness.

  *   ⨀ webmachinelearning/webnn#760<https://github.com/webmachinelearning/webnn/issues/760>
  *   ⛙ webmachinelearning/webnn#830<https://github.com/webmachinelearning/webnn/pull/830>

ℹ️ Core operator set
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#%E2%84%B9%EF%B8%8F-core-operator-set>

Revisit our core operator set effort that aims to identify current primitive gaps by mapping compositional fundamentals (e.g. PyTorch prims, TOSA, StableHLO) to WebNN operators.

  *   ⨀ webmachinelearning/webnn#573<https://github.com/webmachinelearning/webnn/issues/573>
  *   ☰ Machine Learning Operator Mapping - All Raw Operators<https://onedrive.live.com/edit?id=EE82F5C6F06C7371!345450&resid=EE82F5C6F06C7371!345450&ithint=file%2Cxlsx&authkey=!AK8f-RDTleqlLXE&wdo=2&cid=ee82f5c6f06c7371>

ℹ️ Caching mechanism for MLGraph
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#%E2%84%B9%EF%B8%8F-caching-mechanism-for-mlgraph>

Discuss refocused explainer and initial implementation feedback.

  *   ⨀ webmachinelearning/webnn#807<https://github.com/webmachinelearning/webnn/issues/807>
  *   ☰ Explainer: WIP
  *   ⚙️ Initial implementation: shiyi9801/chromium#227<https://github.com/shiyi9801/chromium/pull/227>

ℹ️ Query mechanism for supported devices
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-04-24-wg-agenda.md#%E2%84%B9%EF%B8%8F-query-mechanism-for-supported-devices>

Discuss use cases and requirements from Google Meet, subject to new information.

  *   ⨀ webmachinelearning/webnn#815<https://github.com/webmachinelearning/webnn/issues/815>