15:00:01 <RRSAgent> RRSAgent has joined #webmachinelearning
15:00:01 <RRSAgent> logging to https://www.w3.org/2020/02/06-webmachinelearning-irc
15:00:02 <sukkee> sukkee has joined #webmachinelearning
15:00:06 <Zakim> Zakim has joined #webmachinelearning
15:00:09 <anssik> RRSAgent, make logs public
15:00:14 <Chai> Chai has joined #webmachinelearning
15:00:14 <anssik> Meeting: WebML CG Teleconference – 6 February 2020
15:00:19 <anssik> Chair: Anssi
15:00:23 <anssik> Agenda: https://github.com/webmachinelearning/meetings/blob/master/telcons/2020-02-06-agenda.md
15:00:34 <anssik> Scribe: Anssi
15:00:34 <anssik> scribeNick: anssik
15:00:34 <anssik> Present+ Anssi_Kostiainen
15:00:46 <anssik> Present+ Ningxin_Hu
15:00:49 <anssik> Present+ Chai_Chaoweeraprasit
15:01:06 <anssik> Present+ Ping_Wu
15:01:30 <ningxin_hu> ningxin_hu has joined #webmachinelearning
15:01:31 <eden> eden has joined #webmachinelearning
15:01:46 <ningxin_hu> Present+ Ningxin_Hu
15:01:49 <anssik> Present+ Ganesan_Ramalingam
15:02:29 <anssik> Present+ Eden_(Baidu)
15:03:07 <anssik> RRSAgent, draft minutes v2
15:03:07 <RRSAgent> I have made the request to generate https://www.w3.org/2020/02/06-webmachinelearning-minutes.html anssik
15:03:55 <anssik> anssik: please welcome Baidu folks to the group
15:05:07 <anssik> ... interested in the work this group does,  bringing in Paddle and Paddle-Lite perspective and experience
15:06:42 <anssik> RRSAgent, make logs public
15:06:48 <anssik> RRSAgent, draft minutes v2
15:06:48 <RRSAgent> I have made the request to generate https://www.w3.org/2020/02/06-webmachinelearning-minutes.html anssik
15:06:57 <Jonathan> Jonathan has joined #webmachinelearning
15:07:08 <BaulEun> BaulEun has joined #webmachinelearning
15:07:21 <anssik> TOPIC: conv2d and matMul op definitions
15:07:31 <anssik> Present+ Jonathan_Bingham
15:07:38 <anssik> Present+ Baul_Eun
15:07:56 <anssik> anssik: Flesh out how we proceed adding conv2d and matMul op definitions to the WebNN API spec.
15:08:11 <anssik> ... Proposed work mode to evolve the op definitions iteratively: first land signature and arguments definitions, refine in subsequent PRs based on compatibility study findings.
15:08:23 <anssik> ... Please review and provide feedback on the respective issues prior to the call:
15:08:28 <anssik> -> https://github.com/webmachinelearning/webnn/issues/28 [op compatibility] conv2d #28
15:08:35 <anssik> -> https://github.com/webmachinelearning/webnn/blob/master/op_compatibility/conv2d.md conv2d compat table
15:09:08 <anssik> -> https://github.com/webmachinelearning/webnn/issues/27 [op compatibility] matMul #27
15:09:16 <anssik> anssik: compat table just added Paddle-Lite mapping
15:10:01 <anssik> anssik: any comments?
15:10:10 <anssik> chai: question on versioning
15:10:31 <anssik> ... need to tackle this version question early
15:11:44 <anssik> ... should discuss compat and versioning together when making progress
15:12:37 <anssik> anssik: I expect the editors to make a PR for these ops
15:12:50 <anssik> ningxin_hu: I can take that action, working with Chai, also consider versioning
15:13:19 <anssik> proposed RESOLUTION: Add conv2d and matMul op definitions to WebNN API
15:14:04 <anssik> ningxin_hu: one comment regarding matMul, it lacks compat study findings
15:15:17 <anssik> ... should figure out if the compat study has been useful for conv2d, then scale to matMul other ops
15:16:35 <anssik> Chai: compat table is definitely useful, we do not know all the APIs, maybe needed only for big ops
15:17:26 <anssik> RESOLUTION: Add conv2d and matMul op definitions to WebNN API
15:17:42 <anssik> TOPIC: Revisit inference API to load and run a model
15:17:51 <anssik> anssik: Discuss a proposal for inference API to load and run a model.
15:17:56 <anssik> anssik: quoting the explainer:
15:18:04 <anssik> ... "ML Inference is a proposed web API to take a custom, pre-trained machine learning model in a standard format, and apply it to example data in JavaScript in order to perform inference, like classification, regression, or ranking. The idea is to make it as easy as possible for web developers to use a custom, pre-built machine learning model in their web app, across devices and browsers."
15:19:02 <anssik> -> https://github.com/jbingham/web-ml-inference Web ML Inference Explained
15:19:08 <anssik> -> https://github.com/webmachinelearning/webnn/issues/41 Revisit inference API to load and run a model #41
15:19:14 <anssik> anssik: feedback wanted in issue #41
15:19:41 <anssik> ... Given adequate support, I'll start the process to expand the Community Group's scope per the charter change process to explicitly bring this proposal in scope of this group.
15:22:00 <anssik> Jonathan: motivation, there was a previous issue #3, some folks from Microsoft have interest in web developer focus, aka load & run model
15:22:44 <anssik> ... when I heard from TF.js having reservations with graph API, I was initially disappointed, so this is a level of API that is supported by Google
15:22:53 <anssik> ... complementary to the graph API
15:23:01 <anssik> ... challenge will be the model format
15:23:14 <anssik> ... in explainer, there's more information on it
15:24:11 <anssik> ... question whether operation based model format such WinML, CoreML, TF is the right level
15:25:46 <anssik> ... MLIR is looking at using a lower-than-op-level abstraction, that model format is likely Google's future direction
15:25:58 <anssik> ... MLIR is not ready, so that's an issue
15:26:11 <anssik> ... questions welcome!
15:27:38 <anssik> ningxin_hu: very good topic to discuss
15:27:53 <anssik> ... want to understand the major concerns of the graph API
15:28:23 <anssik> ... building the graph out off ops is a the concern, right?
15:29:37 <anssik> ... in terms of API surface, compilation and execution can be reused, and add load model API
15:31:23 <anssik> TOPIC: Adjourn
15:31:28 <anssik> RRSAgent, draft minutes v2
15:31:28 <RRSAgent> I have made the request to generate https://www.w3.org/2020/02/06-webmachinelearning-minutes.html anssik
15:31:55 <anssik> rrsagent, please leave
15:31:55 <RRSAgent> I see no action items