2023-08-24-wg-agenda.md

WebML WG Teleconference – 24 August 2023 - 14:00-15:00 UTC

See the timezone table ...

San Francisco (U.S.A. - California)	Thu, 24 August 2022	07:00	UTC-7 hours
Boston (U.S.A. - Massachusetts)	Thu, 24 August 2022	10:00	UTC-4 hours
London (United Kingdom - England)	Thu, 24 August 2022	15:00	UTC+1 hours
Berlin (Germany)	Thu, 24 August 2022	16:00	UTC+2 hours
Helsinki (Finland)	Thu, 24 August 2022	17:00	UTC+3 hours
Shanghai (China)	Thu, 24 August 2022	22:00	UTC+8 hours
Tokyo (Japan)	Thu, 24 August 2022	23:00	UTC+9 hours
Corresponding UTC (GMT)	Thu, 24 August 2022	14:00 UTC

Other locations: https://www.timeanddate.com/worldclock/fixedtime.html?iso=20230824T14

Logistics

• Chair: Anssi
• Scribe: ? (see howto, 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/2023/08/24-webmachinelearning-minutes.html

Agenda

ℹ️ Google Chrome team's feedback on WebNN API

Vivek and Joshua B have gathered feedback from various Google teams interested in WebNN and will share their thoughts with the WG for discussion.

• Feedback GH issue for WG discussion: webmachinelearning/webnn#453

ℹ️ WebNN v2: review proposed new ops and data types

Review and discuss the proposed new ops and data types informed by v2 model targets and recent prototyping efforts.

Proposed models:

• Text-to-image: Stable Diffusion unet/VAE/text encoder
• Image segmentation: Segment Everything decoder
• Speech-to-text: Whisper Tiny
• Text-to-text? Summarization, translation, code completion demonstrated by Transformers.js?

Proposed new ops:

• Logical elementwise comparison/selection operations: equal, greater, lesser, logicalNot, elementwiseIf/ternary, greaterOrEqual/lesserOrEqual
• More elementwise unary operations: identity, sqrt, erf (Gauss err func), reciprocal
• Reshaping operations: squeeze, unsqueeze, flattenTo2d
• Data rearrangement operations: expand, gather
• Normalization operations: meanVarianceNormalization
• Index seeking operations: argMin/argMax
• Misc: cast, fillSequence, triangularMatrix, shape
• Others?

Proposed new data types: int64, uint64

Details: webmachinelearning/webnn#375 (comment)

Background material:

• Transformers.js presentation by Joshua Lochner
• Transformer models presentation by Dwayne Robinson
• Transformer-based models and use cases discussion: webmachinelearning/webnn#375
• Proposed v2 ops (open new issues for v2 op proposals)

ℹ️ WebIDL and Infra standard conventions

Decide whether we're ready to merge the zk-conventions-integration integration branch to main.

These changes align the entire specification with modern specification conventions and add stylistic improvements on top that make navigating this specification more delightful experience.

The following resources are made available to the group to assist in this review task:

• Preview of zk-conventions-integration PR
• Diff between main and zk-conventions-integration PR
• Issues addressed and PRs merged to zk-conventions-integration (see "🟡 review feedback" annotations)
• Commits ahead of main

ℹ️ Enhancements

• split() into sizes not widely supported
  • issue: webmachinelearning/webnn#392
• Clarify the restriction for minValue and maxValue of MLClampOptions
  • issue: webmachinelearning/webnn#396