13:55:58 RRSAgent has joined #webmachinelearning 13:55:58 logging to https://www.w3.org/2021/06/10-webmachinelearning-irc 13:56:00 RRSAgent, make logs Public 13:56:01 please title this meeting ("meeting: ..."), anssik 13:56:16 Meeting: WebML WG Teleconference – 10 June 2021 13:56:20 Chair: Anssi 13:56:24 Agenda: https://github.com/webmachinelearning/meetings/blob/master/telcons/2021-06-10-agenda.md 13:57:25 Scribe: Anssi 13:57:29 scribeNick: anssik 13:58:03 Present+ Anssi_Kostiainen 14:00:20 Present+ Ningxin_Hu 14:00:31 Present+ Chai_Chaoweeraprasit 14:00:51 Present+ Dominique_Hazael-Massieux 14:00:53 Present+ Wonsuk_Lee 14:01:34 Present+ Rohit_Hegde 14:01:39 Present+ Anita_Chen 14:01:58 Present+ Rohit_Rehde 14:02:02 ningxin_hu has joined #webmachinelearning 14:02:12 Rama has joined #webmachinelearning 14:02:12 chai has joined #webmachinelearning 14:02:13 wonsuk has joined #webmachinelearning 14:02:43 Rohit_Hegde has joined #webmachinelearning 14:02:53 Present+ Ganesan_Ramalingam 14:03:00 Present+ Rafael_Citron 14:03:05 Present+ Rachel_Yager 14:03:13 Present+ Anita_Chen 14:03:26 Present+ Rachel_Yager 14:03:42 Present+ Rafael_Cintron 14:03:51 RRSAgent, draft minutes 14:03:51 I have made the request to generate https://www.w3.org/2021/06/10-webmachinelearning-minutes.html anssik 14:03:59 Present+ Ningxin_Hu 14:04:21 Present+ Wonsuk_Lee 14:04:58 scribe+ 14:05:12 Anssi: work for Intel, been involved in Web standards & technologies for quite some time 14:05:18 ... started the Machine Learning CG 2 years ago 14:05:35 ... lucky to have this talented people in incubation, now bringing it to formal standardization 14:06:20 Dom: part of W3C technical staff, involved in setting up the Web & ML workshops, also helping chartering the WG 14:06:30 s/shops/shop/ 14:06:59 Anita: I work for Fraunhofer Fokus in Berlin; attended the ML workshop last fall - hear to listen to any updates 14:07:28 Chai: I work for Microsoft, my team built the ML platform for windows - Windows AI technology 14:07:41 ac has joined #webmachinelearning 14:07:43 ping_yu has joined #webmachinelearning 14:07:44 ... I've been involved in the CG 18 months ago, glad to be able to contribute 14:08:08 Rachel: I'm a W3C evangelist and also manage the W3C NY Metro chapter 14:08:19 ... my professional interest is revolving around the big data domain 14:08:41 ... involved in ML algorithms, would like to see how to contribute to the group and use it in future applications 14:09:07 Ningxin: work for intel, based in Shanghai; from the Web Platform engineering team, closing the gap with native 14:09:19 ... I'm co-editing the WebNN API with Chai - happy to work with you all 14:09:54 Ping: from Google, work on tensorflow.js; been working closely with Ningxin on performance and WebGPU backend 14:10:03 ... very glad to be part of this WG to advance ML on the Web 14:10:19 Rafael: I'm a developer on the edge browser team 14:10:37 ... I also represent Microsoft at the WebGL, WebGPU and Immersive Web and Color on the Web groups in W3C 14:11:29 Rohit: Frauenofer Fit in Bonn in Germany; involved in ML algorithms; have been involved in accessibility topics, glad to join this group 14:11:51 Wonsuk: from ETRI, a government research institute in South Korea 14:12:02 ... involved in a couple of projects, incl around AI & data 14:12:23 ... the Web is a major ecosystem in IT - interested in Web ML APIs standardization 14:12:37 Anssi: good split of CG old timers and new comers 14:12:43 Topic: Welcome to the Web Machine Learning Working Group 14:13:00 i/Anssi: work for Intel/Topic: Intros 14:13:14 Anssi: welcome to our first teleconference as a WG! 14:13:24 ... we'll start with a few orientation topics for the group 14:13:29 Subtopic: Where we are, who we are, what are our plans for the next few months 14:13:43 RRSAgent, draft minutes 14:13:43 I have made the request to generate https://www.w3.org/2021/06/10-webmachinelearning-minutes.html dom 14:14:01 Rach has joined #webmachinelearning 14:14:10 Anssi: Where we are: this WG standardizes work incubated in the CG over the past 2 years 14:14:20 ... that's what our charter sets us to do 14:14:43 ... Who we are: we've just heard a round of introductions, with a diverse set of people: browser, hardware, ISV, research & others 14:15:17 ... we are open for anyone to join - if you know people in your communities who would help further this work, please let them know 14:15:46 ... I expect most of the participants from the CG to become part of the WG 14:15:51 ... We're still onboarding new people 14:16:28 ... we're also opened to invited experts - people not affiliated with a Member organization e.g. unaffiliated individuals - we can help with getting them on board 14:16:55 ... in terms of other stakeholders, we will work with experts from so-called "horizontal" groups: security, privacy, accessibility, internationalization, architecture 14:17:20 ... throughout our standardization journey, we will interact with these experts to help ensure our work fulfills the expectations in these spaces 14:18:03 Anssi: with regard to our plans for the next few months, Working Groups are in the business of advancing their specs along the W3C Recommendation track 14:18:18 ... this starts with a "First Public Working Draft" and ends at the final stage, a W3C Recommendation 14:18:26 -> https://www.w3.org/2020/Process-20200915/#rec-track W3C Recommendation Track 14:18:27 ... when it is “done” 14:18:38 ... a substantial amount of work needs to be done between these two stages 14:18:48 ... we're hoping to hit soon the first milestone, the FPWD 14:18:52 ... Our deliverables: 14:19:00 ... 1) Web Neural Network API 14:19:03 ... our key deliverable in this group is the Web Neural Network API 14:19:06 ... 2) Model Loader API (tentative, subject to agreement on model format) 14:19:49 ... The model loader API has a dependency on having an agreed upon model format 14:19:52 ... 3) Ethical issues associated with ML on the Web (Working Group Note aka informative) 14:20:38 ... we want to ask everyone's help in finding people in your communities that can help with these ethical considerations 14:20:40 Geun-Hyung_ has joined #webmachinelearning 14:20:59 Present+ Geun-Hyung 14:22:03 Dom: for CG old timers, the work happens in a similar fashion, a bit more formalism and attention to the broader W3C community 14:22:13 ... some additional scrutiny from the W3C community 14:22:30 ... for WG newcomers, existing background from people working on these topics for the past couple of years 14:22:46 ... with Anssi, we're committed to get you up to speed with this work, and answer any questions you have 14:22:56 ... please do interrupt and let us know! 14:23:21 ... let us know whenever you feel lost and we'll help you out 14:24:01 Subtopic: Introduction to the Web Neural Network API aka WebNN API 14:24:33 RafaelCintron has joined #webmachinelearning 14:24:44 Rachel: is there any information about next phase work in terms of algorithmic approaches? 14:24:47 ... beyond neural networks 14:24:58 q+ for position of other browser vendors 14:25:13 Anssi: our current charter defines our technical scope 14:25:22 ... we have 1 normative deliverable - the WebNN API 14:25:50 ... we will continue to evolve it as part of the WG, improving it to meet the needs of the API customers 14:26:05 ... Another spec we mentioned is the model loader API - not yet ready for WG, but incubated in the CG 14:27:12 ... if we identify promising work beyond that before the end of our current charter, we can always recharter 14:27:22 ... this charter is very focused to push for timely progress 14:27:49 ... if you have proposals in mind for new work, the CG has a process to identify these new ideas 14:27:51 Introduction to the Web Neural Network API aka WebNN API 14:27:55 https://github.com/webmachinelearning/proposals/ 14:27:58 ack wonsuk 14:27:58 wonsuk, you wanted to discuss position of other browser vendors 14:28:08 s/Introduction to the Web Neural Network API aka WebNN API// 14:28:28 Wonsuk: we have Intel, Google & Microsoft - a good starting point in terms of browser vendors 14:28:46 ... but we need to bring more browser vendors into the group 14:28:54 ... do you know the position of other vendors for this activity? 14:29:18 Anssi: the charter was reviewed by the W3C membership, and noone objected to the latest version of it 14:29:35 ... The CG had representatives from all 4 major browser vendors 14:29:51 ... I can't speak for other vendors, but I would expect all of them to join eventually 14:29:51 q+ 14:30:15 ... clearly an important question given that one of our success criteria will be to demonstrate independent implementation of our technologies 14:30:55 ... in the case of our API, the landscape is a bit different from other specs, given that the API may need to build on different OS backends and different supporting hardware 14:31:15 ... once we get to the stage of demonstrating interop, we may build upon that distinct landscape 14:32:06 Dom: currently not Apple or Mozilla have not joined the WG yet 14:32:37 ... we see participation in line with momentum with developers, other stakeholders, it is sometimes a chicken and egg problem 14:33:04 ... we want to make sure the work we develop does work for other vendors 14:33:32 ... as for implementation experience, Anssi described some ways how to demonstrate interop beyond traditional assessment 14:33:49 ... my preference is interop across several axis, including browser vendors 14:34:23 ... the advice I give to WGs for demonstrating interop is how confident other groups are what we have done can be implemented by different people with different perspectives 14:34:49 ... this can be done with a good level of trust and honesty, it is about building the web for its long-term future 14:35:16 ... in the end the Director will review the assessment, but we need to answer the questions in the end, how confident we are the think we build fulfills the promise of the web 14:35:20 q? 14:35:53 ack me 14:36:04 Subtopic: Introduction to the Web Neural Network API aka WebNN API 14:36:24 -> https://github.com/webmachinelearning/webnn/blob/main/explainer.md WebNN Explainer 14:36:25 Anssi: Chai, would you mind giving a high level view of what WebNN is? 14:37:05 Chai: this work has emerged from a necessity: if we look at the ecosystem on windows, mac, mobile phones, 14:37:18 ... we're seeing a lot of hardware innovation to enable high-speed AI 14:37:45 ... in my day job, I interact quite a lot with hardware vendors that are pushing innovations toward better UX 14:38:04 ... but there is a gap between these innovations and what the Web can do today with it 14:38:11 Present+ Geun-Hyung_Kim 14:38:14 ... we see WebNN as the bridge for that gap 14:38:36 RRSAgent, draft minutes 14:38:36 I have made the request to generate https://www.w3.org/2021/06/10-webmachinelearning-minutes.html anssik 14:38:52 ... The goal is to enable native performance & native support in Web browsers without having to deal with the complexity of the underlying OS/hardware/software combination 14:39:28 ... WebNN is trying to achieve this: enabling running ML models across browsers & platforms as seamlessly as possible 14:39:49 ... WebNN is thus an abstraction of ML that can be recognized across industry & platforms 14:40:14 ... there is a lot of overlap and commonality across the innovations that are happening 14:40:31 ... which WebNN tries to represent to provide access to this innovation to Web developers 14:40:33 RRSAgent, draft minutes 14:40:33 I have made the request to generate https://www.w3.org/2021/06/10-webmachinelearning-minutes.html anssik 14:41:17 Anssi: the WebNN Explainer is a great read, a high level view of what the API is - a great starting point before diving into the spec 14:42:28 Subtopic: Demystify the transition process from CG to WG, CG work adoption 14:43:02 Anssi: In short, as Dom already said, we try to make this transition as seamless and transparent for folks who are transitioning from the CG to WG 14:43:29 ... Dom and I will try to take care of as much of the procedural aspects 14:43:48 ... for existing CG participants that haven't joined the WG, now is the time to do to make sure you can continue contributing 14:44:21 ... In terms of transition, we will issue a Call for Consensus today to formally adopt the WebNN API in the Working Group and publish it as a First Public Working Draft, with a targeted release toward the end of June 14:44:30 ... I'll work with Dom to prepare the specification for publication 14:45:31 Dom: First Public Working Draft does not mean WebNN API is ready, it signals the WG agrees this is the right direction of work 14:45:58 ... it also starts Patent Policy considerations, Call for Exclusions for participants 14:46:41 ... to allow publication of the spec on Royalty-Free basis 14:47:43 Anssi: we are confident that the spec meets the bar for FPWD - we even have already exceeded some expectations (e.g. already got some privacy review, TAG review) 14:48:07 Subtopic: Collaboration model with the WebML WG and other W3C groups 14:48:32 Anssi: we are expected to collaborate with the WebML Community Group 14:48:34 s/WG/CG/ 14:48:47 ... I expect most if not all of the key contributors to participate in both groups 14:49:01 ... we're working under the same github organization, sharing a mailing list, etc 14:49:19 ... In terms of other W3C groups, our charter lists other groups we want to work with 14:49:56 ... - GPU for the Web WG - WebGL & WebGPU are closely related technologies to our work, e.g. ML frameworks often have backends on top of WebGL/WebGPU 14:50:05 ... - WebAssembly - another backend in particular for CPU 14:50:21 ... - WebRTC is mentioned given the media processing APIs that would benefit from ML technologies 14:50:40 ... - the Technical Architecture Group - most specifically for their work on ethical design principles 14:51:04 ... This autumn, there will a W3C TPAC meeting and we hope to organize joint meetings with some of these groups 14:51:15 ... to ensure we're building a cohesive Web platform 14:51:19 q? 14:51:35 Subtopic: Meetings and TPAC vF2F 14:52:00 piyush has joined #webmachinelearning 14:52:02 Anssi: The CG had been using this biweekly slot we're using today - Thursdays at 1400 UTC 14:52:17 ... we'll run a poll to see if that timeslot is still the best option 14:52:35 ... we have a very wide geographic distribution, so there are only a limiter number of possibilities that can work 14:52:43 ... W3C runs a general yearly meeting called TPAC 14:52:53 ... this year again, it will be fully virtual 14:53:03 q? 14:53:11 ... it's an opportunity for joint meetings with other groups, breakouts on dedicated topics 14:53:17 Subtopic: Planned first key WG milestones: 14:53:52 Anssi: After this call, I'll start a call for consensus (CfC) to adopt WebNN API to publish as a FPWD, with a target date of June 24 14:54:12 ... a CfC is a way to check the group and the community has no concerns with that adoption 14:54:17 -> https://webmachinelearning.github.io/ Web Machine Learning website 14:54:17 -> https://www.w3.org/groups/wg/webmachinelearning W3C WG home 14:54:17 -> https://www.w3.org/groups/wg/webmachinelearning/instructions Instructions for joining 14:54:17 -> https://webmachinelearning.github.io/webnn/ WebNN API spec 14:54:17 -> https://github.com/webmachinelearning/webnn/blob/master/explainer.md WebNN API explainer 14:54:44 Topic: WebNN API - new proposals 14:54:57 Anssi: for new people, this should be a good way to see how we handle our day-to-day work 14:55:09 -> https://github.com/webmachinelearning/webnn/issues/156 Issue #156 14:55:15 -> https://github.com/webmachinelearning/webnn/pull/174 PR #174 14:55:28 Anssi: this particular item is a follow up item from a previous CG call 14:55:45 ... Ningxin submitted a pull request for sync build & compute and require pre-allocated output buffers in compute 14:56:16 ... This is the op-level pre-allocation use case described in pull request #174, based on the issue describe in #156 14:56:27 ... I would like to call Ningxin to describe the 3 main changes in this pull request 14:56:43 ... The pull request has been reviewed already quite extensively, hopefully can be merged after the call 14:57:08 Anssi: Ningxin to talk us through the three key changes: 14:57:16 ... 1) Change MLGraphBuilder.build and MLGraph.compute to sync API. 14:57:16 ... 2) Change MLGraph.compute to require pre-allocated output buffers. 14:57:16 ... 3) Simplify the the input / output resources binding by MLResource. Leave MLInput only for dynamic input shape case. 14:57:17 Ningxin: the PR is built from a previous PR where we had lots of good discussions 14:57:24 ... this is focusing on the WASM use case and the Operation-level use case 14:57:30 RRSAgent, draft minutes 14:57:30 I have made the request to generate https://www.w3.org/2021/06/10-webmachinelearning-minutes.html anssik 14:57:42 ... during the discussion, we found this is quite useful, not only for op-level execution 14:58:15 ... but also for other use cases - e.g. in tensorflow-lite WASM 14:58:32 ... the feature in this PR will be useful also for graph-level execution 14:58:44 ... The PR includes 2 major changes: 14:59:07 ... change the MLGraphBuilder.build and MLGraph.compute APIs as synchronous APIs 14:59:47 ... the motivation is that synchronous APIs are easier to use, and more importantly, the WASM APIs are expecting a sync API 15:00:27 ... Ping mentioned in #156 - tensorflow.js shared a way to avoid blocking the main thread with a sync API using Web workers 15:00:42 ... The second change is to require to pre-allocate output buffers for the compute API 15:00:59 ... This is also based on the needs from ML frameworks 15:01:20 ... most of the WASM libraries will prepare output buffers and will expect the compute APIs to put the output in the WASM memory 15:01:26 Present+ Piyush_Chauhan 15:01:36 RRSAgent, draft minutes 15:01:36 I have made the request to generate https://www.w3.org/2021/06/10-webmachinelearning-minutes.html anssik 15:01:37 ... for the GPU, creating GPU resources on the fly is very expensive - so it's also good for the GPU usage 15:02:48 ... Another change is to simplify the resources binding 15:03:26 Anssi: Rafael, any thought? We've had reviews from Ping & Chai 15:03:44 Rafael: haven't looked in details into it, but reading the discussion, it looks good to me 15:04:08 Anssi: OK, shall we it merge it then? 15:04:23 Rafael: no concern for me; we can fix problems later if needed 15:04:30 RESOLVED: Merge https://github.com/webmachinelearning/webnn/pull/174 15:05:23 Topic: AOB 15:05:47 Anssi: we will defer the remaining items to our next call - schedule to be confirmed given starting vacation schedules 15:05:48 q? 15:06:26 Anssi: Thank you all for joining today! 15:06:53 RRSAgent, draft minutes 15:06:53 I have made the request to generate https://www.w3.org/2021/06/10-webmachinelearning-minutes.html dom 15:07:00 RRSAgent, make log public 15:50:57 join_subline has joined #webmachinelearning 17:35:09 Zakim has left #webmachinelearning