WebML WG Teleconference – 17 October 2024 - 14:00-15:00 UTC

Latest version: https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md



WebML WG Teleconference – 17 October 2024 - 14:00-15:00 UTC
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#webml-wg-teleconference--17-october-2024---1400-1500-utc>
See the timezone table ...

Logistics
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#logistics>

  *   Chair: Anssi
  *   Scribe: ? (see howto<https://github.com/webmachinelearning/meetings/blob/main/scribe-howto.md>, volunteers welcome!)
  *   IRC: irc://irc.w3.org:6667/#webmachinelearning

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

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

  *   Minutes: https://www.w3.org/2024/10/17-webmachinelearning-minutes.html


Agenda
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#agenda>

Note

We'll resume our bi-weeklies with a refresher on selected resolutions and proposals from F2F, and discuss our next steps.

ℹ️ WebNN Operator Update Wave 3
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#%E2%84%B9%EF%B8%8F-webnn-operator-update-wave-3>


🟢 Resolution: Update spec with Wave 3 operators, initiate int4/uint4 wide review.

  *   📁 Slides<https://lists.w3.org/Archives/Public/www-archive/2024Sep/att-0014/WebNN_Operator_Update_Wave_3.pdf>
  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#798f>

ℹ️ Quantization and dequantization (QDQ)


🟢 Resolution: Add QDQ operators for int8 and int4 and consolidate #93 #128 #623.

  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#2842>

ℹ️ Platform capability detection


✋ Proposal: Collect real-world feedback from opSupportLimits() usage to inform baseline and max limits.

  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#b5cc>
  *   ⨀ webmachinelearning/webnn#463<https://github.com/webmachinelearning/webnn/issues/463>
  *   ⛙ webmachinelearning/webnn#755<https://github.com/webmachinelearning/webnn/pull/755>

ℹ️ Device selection abstractions


✋ Proposal: Draft a spec PR to remove MLDeviceType, keep MLPowerPreference. Gauge prototyping interest and impact to frameworks?

  *   📁 Slides<https://lists.w3.org/Archives/Public/www-archive/2024Sep/att-0006/MLDeviceType.pdf>
  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#b5c1>

ℹ️ Google Chrome feedback revisited
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#%E2%84%B9%EF%B8%8F-google-chrome-feedback-revisited>


🟢 Resolution: Close Chrome feedback #453<https://github.com/webmachinelearning/webnn/issues/453> as completed.

  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#899d>

ℹ️ Interop issues across different backends
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#%E2%84%B9%EF%B8%8F-interop-issues-across-different-backends>


✋ Proposal: Revisit interop issues<https://github.com/webmachinelearning/webnn/issues?q=is%3Aissue+is%3Aopen+label%3Ainterop> e.g. remove pooling's rounding direction from MLPool2dOptions to close #324, decide between clamp or "ignore the index" approach to close #486.

  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#7320>

ℹ️ Core operator set
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#%E2%84%B9%EF%B8%8F-core-operator-set>


✋ Proposal: Document requirements for adding "new core op" and "non-core op", (consider e.g. TOSA, MLIR linalg), categorize ops in spec.

  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#23a8>
  *   ⨀ webmachinelearning/webnn#573<https://github.com/webmachinelearning/webnn/issues/573>

ℹ️ MLTensor
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#%E2%84%B9%EF%B8%8F-mltensor>


✋ Proposal: Merge explainer PR with an agreement MLDeviceType changes to impact buffer allocation.

  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#e5b7>
  *   ⛙ webmachinelearning/webnn#754<https://github.com/webmachinelearning/webnn/pull/754>

ℹ️ Wide review: TAG
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#%E2%84%B9%EF%B8%8F-wide-review-tag>


🟢 Resolution: Add resource contention considerations to the spec to address TAG review feedback.

  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#cdd5>
  *   ⛙ webmachinelearning/webnn#765<https://github.com/webmachinelearning/webnn/pull/765>

Tensor primitives
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#tensor-primitives>


✋ Proposal: Continue explore authoring high-level ops with tensor primitives.

  *   📁 Slides<https://lists.w3.org/Archives/Public/www-archive/2024Sep/att-0007/Tensor_Primitive_Ops_Proposal_-_TPAC.pdf>
  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#b039>

Translation and Prompt APIs
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2024-10-17-wg-agenda.md#translation-and-prompt-apis>


✋ Proposal: Solicit input on whether to adopt these APIs as new deliverables in the WG's Charter 2025->

  *   📁 Slides<https://lists.w3.org/Archives/Public/www-archive/2024Sep/att-0008/TPAC_2024_Built-in_AI_APIs.pdf>
  *   ✍️ F2F minutes<https://www.w3.org/2024/09/23-webmachinelearning-minutes.html#20fb>