IRC log of webmachinelearning on 2023-03-30

Timestamps are in UTC.

14:00:18 [RRSAgent]

RRSAgent has joined #webmachinelearning

14:00:22 [RRSAgent]

logging to https://www.w3.org/2023/03/30-webmachinelearning-irc

14:00:22 [Zakim]

RRSAgent, make logs Public

14:00:23 [Zakim]

please title this meeting ("meeting: ..."), anssik

14:00:23 [ningxin_hu]

ningxin_hu has joined #webmachinelearning

14:00:23 [anssik]

Meeting: WebML WG Teleconference – 30 March 2023

14:00:29 [anssik]

Chair: Anssi

14:00:34 [anssik]

Agenda: https://github.com/webmachinelearning/meetings/blob/main/telcons/2023-03-30-wg-agenda.md

14:00:53 [anssik]

Scribe: Anssi

14:01:02 [anssik]

scribeNick: anssik

14:01:04 [zkis]

zkis has joined #webmachinelearning

14:01:11 [anssik]

Regrets+ Dominique_Hazael-Massieux

14:01:34 [anssik]

ghurlbot, this is webmachinelearning/webnn

14:01:34 [ghurlbot]

anssik, OK.

14:01:36 [anssik]

Present+ Anssi_Kostiainen

14:01:41 [anssik]

Present+ Ningxin_Hu

14:01:45 [anssik]

Present+ Zoltan_Kis

14:02:26 [anssik]

Present+ Chai_Chaoweeraprasit

14:04:36 [anssik]

Present+ Rafael_Cintron

14:04:44 [anssik]

RRSAgent, draft minutes

14:04:45 [RRSAgent]

I have made the request to generate https://www.w3.org/2023/03/30-webmachinelearning-minutes.html anssik

14:04:52 [RafaelCintron]

RafaelCintron has joined #webmachinelearning

14:05:11 [chai]

chai has joined #webmachinelearning

14:05:46 [anssik]

Topic: Web Neural Network API W3C Candidate Recommendation published

14:05:56 [anssik]

-> Web Neural Network API CR, 30 March 2023 https://www.w3.org/TR/2023/CR-webnn-20230330/

14:06:10 [anssik]

-> W3C invites implementations of Web Neural Network API https://www.w3.org/blog/news/archives/9868

14:06:29 [anssik]

-> ▶️ Clapping Sound Effects https://www.youtube.com/watch?v=jDOrc8FmDy4

14:06:33 [chai]

🎉

14:07:16 [ningxin_hu]

Cheers!

14:07:21 [anssik]

anssik: Thank you to all the WG participants for your hard work that made this publication possible!

14:07:37 [anssik]

... this has been a collaboration between lead implementers (thank folks at Microsoft and Google!),

14:08:00 [anssik]

... interoperability at its heart (thanks spec and wpt contributors across companies!),

14:08:24 [anssik]

... with feedback from early adopters, web developers, other WGs (thanks WebGPU WG, WebRTC WG!) informing the effort.

14:08:46 [anssik]

... Now is a time to celebrate. It is no small feat to hit CR so it is OK to be a little proud :)

14:08:59 [anssik]

... Then, about the next steps.

14:09:21 [anssik]

... As a concrete first next step, we should configure Echidna to again allow publications in this post-CR world

14:09:33 [anssik]

... This commit disabled publications temporarily https://github.com/webmachinelearning/webnn/commit/00d3cdd721e7918206808fcc318b27cbf714b4f5

14:09:50 [anssik]

... I assume Dom can help configure Echidna the right way

14:10:10 [anssik]

... about the new spec maturity levels on the W3C Rec Track in this post-CR state, my take:

14:11:06 [anssik]

... for "corrections that do not affect conformance" incl. editorial enhancements we publish a CR Draft:

14:11:11 [anssik]

-> CR Draft https://www.w3.org/2021/Process-20211102/#publishing-crud

14:11:30 [anssik]

... for "corrections that may [...] affect conformance" or add "new features" we published (a new) CR Snapshot

14:11:35 [anssik]

-> CR Snapshot https://www.w3.org/2021/Process-20211102/#publishing-crrs

14:11:52 [anssik]

... OTOH, the W3C Process guidance is "a Candidate Recommendation Snapshot should not be published more often than approximately once every 6 months."

14:12:23 [anssik]

... so my understanding is we should set Echidna to auto-publish CR Drafts similarly to WDs earlier and should put a clock running and ~6 months from now publish a new CR Snapshot

14:12:57 [anssik]

... and prior to CR Snaphot we should seek wide review for any substantive changes made since this initial CR. We only request review for the delta, so more light weight.

14:13:09 [anssik]

RRSAgent, draft minutes

14:13:11 [RRSAgent]

I have made the request to generate https://www.w3.org/2023/03/30-webmachinelearning-minutes.html anssik

14:14:30 [anssik]

q?

14:14:44 [anssik]

Topic: WebNN - WebIDL and Infra standard conventions

14:15:41 [anssik]

anssik: Zoltan has worked on editorial enhancements to align the spec with WebIDL and Infra standard conventions. Now that the CR is out we plan to start integrating these

14:16:09 [anssik]

Subtopic: The constant() method steps

14:16:14 [anssik]

-> Spec: Create an operand for a graph constant https://www.w3.org/TR/webnn/#dom-mlgraphbuilder-constant

14:16:19 [anssik]

-> Spec: Create a single-value operand from the specified number of the specified type https://www.w3.org/TR/webnn/#dom-mlgraphbuilder-constant-value-type

14:16:23 [anssik]

... PR #365

14:16:24 [ghurlbot]

https://github.com/webmachinelearning/webnn/issues/365 -> Pull Request 365 Add the constant() method steps. (zolkis)

14:17:07 [anssik]

Zoltan: thanks Ningxin for reviewing this PR and improving it!

14:17:26 [anssik]

... ready for review, needs two approvals

14:18:01 [anssik]

... constants will add two forms, dependency on MLOperand, MLActivation not needed here

14:18:34 [anssik]

... completed algorithms for all the ops in the spec and have ~50 PRs ready

14:18:49 [anssik]

... bikeshed builds OK, all dependencies addressed

14:20:34 [anssik]

[ Zoltan sharing a preview of the improved constant() ]

14:21:16 [anssik]

Zoltan: we keep track of platform objects for given operands and track reference to operand

14:21:32 [anssik]

... we should add arch diagram to the spec to help explain this

14:22:18 [anssik]

... not detailing the algorithms per se e.g. for conv2d just say what to do with inputs and how to get outputs, that is what we want to accomplish with this

14:22:40 [anssik]

... some algorithms are long such as concat()

14:22:59 [anssik]

... we can factor out parts of the algorithms as needed

14:23:20 [anssik]

... lstm() is more challenging one

14:23:42 [anssik]

... asking for feedback how to split this work to make it easy for reviewers, one PR per op or multiple ops per PR?

14:23:51 [anssik]

anssik: thanks Zoltan!

14:23:54 [anssik]

q?

14:27:07 [chai]

q+

14:27:11 [anssik]

ack chai

14:27:33 [anssik]

chai: my only recommendation for this type of change to get it in sooner rather than later

14:29:12 [anssik]

... because the change affects the overall style and structure of the document, we should get this in soon before we add more content so new changes can follow this modernized convention

14:30:02 [anssik]

Zoltan: the PRs submitted minimize changes so some of the visual changes presented are not in PRs

14:30:35 [anssik]

... Chai, do you prefer to add visual enhancements first?

14:32:13 [anssik]

Chai: you should look at the changes in two categories: 1) any styling or structure changes that affect every op, do this first 2) enhancing each op that need to be enhanced, this you can group as another category

14:32:43 [anssik]

... otherwise people who want to add new PRs will be blocked on not conforming with the latest styling or structural changes

14:33:12 [anssik]

Zoltan: if I change one method at a time e.g. constant() it does not block adding another op with different styling

14:33:30 [anssik]

Chai: you should keep presentation consistent within the entire specification

14:33:47 [anssik]

... if new ops are added, then you'd need to re-style those new ops added

14:33:49 [ningxin_hu]

q+

14:33:59 [anssik]

... prefer styling to be integrated as soon as possible

14:34:21 [ningxin_hu]

+1 to make the structure change first

14:34:28 [anssik]

Zoltan: issue is if no algorithms or internal slots are defined then we'd have empty algorithm and internal slot boxes

14:34:55 [anssik]

Chai: if those are optional, then that can wait, I'm waiting about style changes that apply broadly

14:35:05 [ningxin_hu]

the algorithm and internal slots to be added could be placeholders there

14:35:20 [anssik]

... anything that is a style change should land as soon as possible

14:36:27 [anssik]

q?

14:36:49 [anssik]

ack ningxin_hu

14:37:04 [anssik]

ningxin_hu: thanks Zoltan for this great work, I like the updated visuals!

14:37:33 [anssik]

... I vote for Chai's opinion, make structure changes first, layout each method in this style you have

14:38:58 [anssik]

... I like the algorithm boxes, if not defined yet suggest to make them placeholders

14:39:37 [anssik]

... e.g. lstm could be many PRs, for style changes I recommend land it in one PR

14:39:48 [anssik]

Zoltan: do you like algorithm box like presented?

14:40:34 [anssik]

Present+ Dom

14:40:45 [anssik]

Dom: prefer to have steps name inside the box

14:41:20 [anssik]

Zoltan: Arguments and returns now in prose, options inside the arguments and returns section

14:41:34 [anssik]

... should I move those out to proper definitions of dictionaries?

14:42:04 [anssik]

... this would be for most algorithms the whole thing, without this I will just change with the styling

14:46:52 [anssik]

ningxin_hu: I think moving options out improves readability

14:47:28 [anssik]

... options are kind of a type definition, moving it out into a separate section I think will improve it and can be reused by other adjesent ops

14:47:45 [anssik]

s/adjesent/adjecent

14:48:08 [chai]

q+

14:48:15 [anssik]

Zoltan: I can prep a PR for review and import it to other branches I have

14:48:48 [ningxin_hu]

That's a significant effort, thanks so much Zoltan

14:49:38 [anssik]

ack chai

14:49:57 [anssik]

Chai: thanks for this, it is great!

14:50:10 [anssik]

... on the algorithm section, some can be auto-generated perhaps

14:50:26 [anssik]

Zoltan: I tried it actually

14:50:48 [anssik]

Chai: we're going to add more ops in the future, if we need to add this boilerplate and stay consistent it is a huge amount of work

14:51:02 [anssik]

... it depends on how complex the op is going to be

14:51:09 [anssik]

... very hard to keep them consistent

14:51:34 [anssik]

... need to factor things out and generate common prose

14:51:47 [anssik]

Zoltan: we should refactor things out and reuse whenever possible

14:51:55 [anssik]

... now I think you see the big picture

14:53:05 [anssik]

... this will still be pretty manual process, so need to see what can be factored out, we should do automation exploration

14:53:26 [anssik]

Chai: if the doc is in transient state or inconsistent it will be hard for others to add new things

14:53:53 [dom]

[overtime, we can also add CI jobs to enforce consistency - I've done that in a number of cases in WebRTC]

14:53:57 [anssik]

Zoltan: once we get these PRs in them can borrow updated conventions

14:54:02 [anssik]

q?

14:54:07 [dom]

q+

14:54:46 [anssik]

Chai: optional visibility with <details> would be nice

14:55:04 [anssik]

... original concern was it should be easy for editors to add new stuff

14:55:05 [anssik]

ack dom

14:55:28 [anssik]

dom: I have experiences with specs that add consistency checks as CI jobs once we know what we expect as input and what as output

14:55:47 [anssik]

... Zoltan is willing to do the manual work now, let's try to get that done as soon as possible to not block the rest of the work

14:55:57 [anssik]

... once we have that base established we look into automation opportunities

14:56:20 [anssik]

Zoltan: if we hide sections, is it better to hide algorithm, validation etc.

14:56:41 [anssik]

dom: parameter hiding sounds good, it is the most boring part

14:57:11 [anssik]

Zoltan: need to check if we can make the whole box disappear and auto-hide it

14:57:25 [anssik]

... I got the feedback, we'll look at it next time with an updated PR

14:57:27 [anssik]

q?

14:58:24 [anssik]

Zoltan: I will ping you on an updated PR with visual enhancements and ping you all

14:58:27 [anssik]

q?

14:59:26 [anssik]

Zoltan: PR #365 is ready and welcomes more reviews!

14:59:27 [ghurlbot]

https://github.com/webmachinelearning/webnn/issues/365 -> Pull Request 365 Add the constant() method steps. (zolkis)

15:00:10 [chai]

i'll take a look

15:01:24 [anssik]

anssik: thanks Chai!

15:03:21 [anssik]

RRSAgent, draft minutes

15:03:22 [RRSAgent]

I have made the request to generate https://www.w3.org/2023/03/30-webmachinelearning-minutes.html anssik

17:29:09 [Zakim]

Zakim has left #webmachinelearning