WebML WG Teleconference – 19 December 2024 - 15:00-16:00 UTC

Latest version: https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md


WebML WG Teleconference – 19 December 2024 - 15:00-16:00 UTC
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#webml-wg-teleconference--19-december-2024---1500-1600-utc>
See the timezone table …


Logistics
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#logistics>

  *   Chair: 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/2024/12/19-webmachinelearning-minutes.html


Agenda
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#agenda>

ℹ️ WebML Working Group Charter update
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#%E2%84%B9%EF%B8%8F-webml-working-group-charter-update>


Working Group charter is up for renewal in 2025. We'll review the current charter, discuss and solicit input on proposed changes, triage open charter issues, and confirm the timeline. Please open issues for any proposed changes or questions.


References:

  *   ☰ WG Charter 2023-2024: https://www.w3.org/2023/04/web-machine-learning-charter.html

  *   ⨀ WG Charter open issues: https://github.com/w3c/machine-learning-charter/issues


ℹ️ WebNN Operator Update Wave 3
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#%E2%84%B9%EF%B8%8F-webnn-operator-update-wave-3>


🎁 Subject to readiness.


Review spec PR for Wave 3 operators and new data types, address any open questions.


  *   Data reorganization: gatherElements, scatterElements, gatherND, scatterND, tile
  *   Elementwise unary: sign
  *   Elementwise binary: logicalAnd, logicalOr, logicalXor, (notEqual)
  *   Elementwise trinary: dequantizeLinear, quantizeLinear
  *   New data types: uint4, int4


References:

  *   ⨀ webmachinelearning/webnn#375<https://github.com/webmachinelearning/webnn/issues/375>
  *   ⛙ PR to be added
  *   📁 22 Sep 2024 TPAC presentation<https://lists.w3.org/Archives/Public/www-archive/2024Sep/att-0014/WebNN_Operator_Update_Wave_3.pdf>
  *   ✅ 17 Oct 2024 resolution<https://www.w3.org/2024/10/17-webmachinelearning-minutes.html#1152>

ℹ️ Device selection abstractions update
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#%E2%84%B9%EF%B8%8F-device-selection-abstractions-update>


Review updated proposal, new use cases, known implementation limitations, and address open questions.


References:

  *   ⛙ webmachinelearning/webnn#784<https://github.com/webmachinelearning/webnn/pull/784>

ℹ️ Core op set: MLIR Linalg findings revisited
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#%E2%84%B9%EF%B8%8F-core-op-set-mlir-linalg-findings-revisited>


Review and discuss the identified gaps / proposed additions to the WebNN op set informed by MLIR Linalg:


  *   1-D convolution with no channels conv_1d<https://mlir.llvm.org/docs/Dialects/Linalg/#linalgconv_1d-linalgconv1dop>
  *   3-D convolution with no channels conv_3d<https://mlir.llvm.org/docs/Dialects/Linalg/#linalgconv_3d-linalgconv3dop>
  *   Fill output with random numbers fill_rng_2d<https://mlir.llvm.org/docs/Dialects/Linalg/#linalgfill_rng_2d-linalgfillrng2dop>
  *   Sum pooling pooling_nchw_sum<https://mlir.llvm.org/docs/Dialects/Linalg/#linalgpooling_nchw_sum-linalgpoolingnchwsumop>
  *   Min pooling pooling_nhwc_min<https://mlir.llvm.org/docs/Dialects/Linalg/#linalgpooling_nhwc_min-linalgpoolingnhwcminop>
  *   Round(x) elementwise round<https://mlir.llvm.org/docs/Dialects/Linalg/#linalground-linalgroundop>


References:

  *   ☰ WebNN-to-ONNX/TOSA/MLIR Linalg op mapping table<https://onedrive.live.com/view.aspx?resid=EE82F5C6F06C7371%21345450&authkey=!AK8f-RDTleqlLXE>

ℹ️ MLTensor: open issues
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#%E2%84%B9%EF%B8%8F-mltensor-open-issues>


Continue discussion on timeline, non-fatal errors, other topical issues that need solutions to enable MLTensor key scenarios: buffer reuse, chained inference, resource management, WebGPU interop.

  *   ☰ mltensor-explainer.md<https://github.com/webmachinelearning/webnn/blob/main/mltensor-explainer.md>


References:

  *   ⨀ Timelines webmachinelearning/webnn#529<https://github.com/webmachinelearning/webnn/issues/529>
  *   ⨀ Non-fatal errors webmachinelearning/webnn#778<https://github.com/webmachinelearning/webnn/issues/778>
  *   ⨀ Multiple devices webmachinelearning/webnn#350<https://github.com/webmachinelearning/webnn/issues/350>
  *   ⨀ Ctor vs factory webmachinelearning/webnn#697<https://github.com/webmachinelearning/webnn/issues/697>

ℹ️ Get to know Task-specific APIs and Prompt API
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-12-19-wg-agenda.md#%E2%84%B9%EF%B8%8F-get-to-know-task-specific-apis-and-prompt-api>


Our companion Community Group now incubates selected task-specific APIs to enable reuse of the built-in models that are distributed as part of the browser or the underlying software platform.

We discuss a few top-of-mind open issues for each API to chart the problem space. Bring your own issues!


References:

  *   🆕 Incubations landing page<https://webmachinelearning.github.io/incubations/>
  *   ☰ Translator and Language Detector APIs (explainer<https://github.com/webmachinelearning/translation-api/>, issues<https://github.com/webmachinelearning/translation-api/issues>)
  *   ☰ Writing Assistance APIs (explainer<https://github.com/webmachinelearning/writing-assistance-apis>, spec<https://webmachinelearning.github.io/writing-assistance-apis/>, issues<https://github.com/webmachinelearning/writing-assistance-apis/issues>)
  *   ☰ Prompt API (explainer<https://github.com/webmachinelearning/prompt-api>, issues<https://github.com/webmachinelearning/prompt-api/issues>)

🎄 Happy Holidays!