14:01:32 <RRSAgent> RRSAgent has joined #webmachinelearning
14:01:32 <RRSAgent> logging to https://www.w3.org/2020/10/01-webmachinelearning-irc
14:01:37 <Zakim> Zakim has joined #webmachinelearning
14:01:42 <anssik> RRSAgent, make logs public
14:01:46 <anssik> Meeting: WebML CG Teleconference – 1 October 2020
14:01:51 <anssik> Chair: Anssi
14:01:55 <anssik> Agenda: https://github.com/webmachinelearning/meetings/blob/master/telcons/2020-10-01-agenda.md
14:02:04 <anssik> Scribe: Anssi
14:02:05 <anssik> scribeNick: anssik
14:02:09 <anssik> Present+ Anssi_Kostiainen
14:02:20 <anssik> Present+ Ganesan_Ramalingam
14:02:24 <ningxin_hu> Present+ Ningxin_Hu
14:02:25 <anssik> Present+ Chai_Chaoweeraprasit
14:02:40 <anssik> Present+ Rafael_Cintron
14:02:49 <anssik> RRSAgent, draft minutes v2
14:02:49 <RRSAgent> I have made the request to generate https://www.w3.org/2020/10/01-webmachinelearning-minutes.html anssik
14:02:56 <anssik> TOPIC: Model Execution API
14:03:49 <anssik> anssik: chai noted by introducing the notion of topology in the API, it acts as an agent to the context that owns the transient state of the topology being constructed.
14:03:51 <anssik> ... this allows for more efficient garbage collection of nodes
14:04:06 <Chai> Chai has joined #webmachinelearning
14:04:14 <anssik> anssik: Rama mentions how the proposed API might allow the API to act as both a graph builder and an eager API that evals ops as they are constructed
14:04:24 <anssik> anssik: Ningxin proposed an 'eager' arg for getNeuralNetworkContext
14:04:33 <anssik> anssik: proposal to track eager model in a separate issue
14:04:39 <anssik> ... considered an implementation variant of the same context interface
14:05:07 <anssik> -> https://github.com/webmachinelearning/webnn/issues/87 Model Execution API #87
14:06:08 <anssik> Chai: this is a long discussion, chatting with Ningxin offline that we should handle with this in separate issues
14:06:32 <anssik> ... secondly, I have pushed a PR out that hopefully addresses the issue
14:07:07 <anssik> -> https://github.com/webmachinelearning/webnn/pull/94 Update model execution API PR#94
14:07:12 <RafaelCintron> RafaelCintron has joined #webmachinelearning
14:07:50 <anssik> Chai: key point around how we separate different states, neuralnetworkcontext should keep the state, model builder is the place where the implementation can keep the state that is specific to the graph building session
14:08:19 <anssik> ... with proper separation of states, GC easier, once you produce a model you no longer need them and they came be CG'd
14:08:51 <anssik> Chai: I think PR #94 ffixes this issue #87
14:09:02 <anssik> ... eager has a separate issue so be can piggy-pack on that one
14:10:16 <anssik> -> https://github.com/webmachinelearning/webnn/issues/11 Executing operations (issue #11)
14:10:25 <anssik> anssik: issue #11 for eager mode follow ups
14:11:57 <ningxin_hu> https://github.com/webmachinelearning/webnn/pull/22
14:12:27 <Chai> q+
14:12:34 <anssik> ack Chai
14:12:41 <ningxin_hu> and https://github.com/webmachinelearning/webnn/issues/16
14:15:30 <ping_yu> ping_yu has joined #webmachinelearning
14:15:46 <anssik> ningxin_hu: PR #94 LGTM, apart from a separate issue for scalar tensor
14:16:34 <anssik> Present+ Ping_Yu
14:17:04 <ping_yu> morning, I will do that
14:17:56 <anssik> anssik: anything else to discuss for issue #87?
14:18:06 <anssik> TOPIC: WebNN code editor
14:18:21 <anssik> s/TOPIC: WebNN code editor//
14:18:33 <anssik> ningxin_hu: a comment re model loader, may need Jonathan's input on that one
14:19:20 <anssik> Rama: just a minor questions, probably a different issue re model loader
14:19:40 <anssik> ... the only question is whether we need indirection to load two different models over on a single call
14:19:57 <anssik> q?
14:21:35 <anssik> ... if having a separate model loader makes no sense, we could have the WebNN and Model Loader under the same API
14:22:20 <zkis> The model loader API could have an `options` optional arg for specifying constraints
14:23:16 <anssik> Chai: I don't see a drawback with Rama's proposal, but we need to think this a bit more and open a separate issue for merge proposals, with input from Jonathan
14:24:41 <anssik> anssik: should understand the scope of WebNN API at the time of WG transition with is expected to happen by the end of the year
14:26:51 <anssik> https://github.com/webmachinelearning/webnn/issues/90
14:27:09 <anssik> zkis: for model loader, options arg would make sense, to be able to pass constraints
14:27:51 <anssik> anssik: API ergonomics improvement?
14:28:17 <anssik> q?
14:28:27 <anssik> TOPIC: WebNN code editor
14:28:35 <anssik> anssik: Review and solicit feedback on the WebNN code editor:
14:28:39 <anssik> -> https://github.com/webmachinelearning/webnn-samples/pull/11 Add the code examples (PR #1)
14:28:45 <anssik> -> https://github.com/webmachinelearning/webnn-samples/issues/4 Add the WebNN code example (issue #4)
14:28:49 <anssik> -> https://huningxin.github.io/webnn-samples/code/ Code editor (in staging)
14:29:34 <anssik> ningxin_hu: this feature is inspired by TF.js docs, discussed with Ping already
14:30:12 <anssik> ... we may want to run the example code in our spec with this feature
14:30:44 <anssik> ... also make the example code in the spec editable
14:31:22 <anssik> ... by default read-only, can turn into edit mode to turn the code snippet into editable code
14:32:12 <anssik> ... considering adding a mechanism to switch between examples from within the code editor
14:32:52 <Chai> this is an excellent idea. it'll help with making sure that the samples and the spec stay consistent.
14:33:00 <anssik> ... later we can add graph builder LeNet example
14:33:12 <ping_yu> looks great and very useful
14:34:07 <ping_yu> It would be good to integrate this with the spec doc
14:34:32 <anssik> anssik: everyone is in agreement this code editor is a great feature, LGTM!
14:36:01 <anssik> TOPIC: WebNN API explainer
14:36:10 <anssik> anssik: TAG review readiness depends on the explainer document
14:36:27 <anssik> -> https://github.com/webmachinelearning/webnn/blob/master/explainer.md WebNN explainer
14:36:40 <anssik> anssik: Discuss how to approach the following explainer sections:
14:36:45 <anssik> ... Key scenarios
14:36:50 <anssik> ... Detailed design discussion
14:36:55 <anssik> ... Considered alternatives
14:39:17 <anssik> -> https://github.com/w3ctag/w3ctag.github.io/blob/master/explainers.md Explainer explainer
14:40:36 <anssik> -> https://github.com/w3ctag/design-reviews/issues TAG spec reviews repo
14:40:53 <Chai> i can take a peak and see how to help. is there any timeline on this?
14:41:11 <anssik> anssik: no timeline per se
14:41:52 <zkis> Sangwhan can probably proxy this spec in the TAG, but IMHO other technical members would wonder if the API could be made more JS-like
14:43:12 <anssik> zkis: Kenneth of TAG provided feedback on making the API more JS-like
14:45:20 <ningxin_hu> i can help as well
14:46:08 <zkis> https://w3ctag.github.io/design-principles/
14:46:11 <anssik> https://github.com/webmachinelearning/webnn/issues/18
14:46:16 <anssik> https://github.com/webmachinelearning/webnn/issues/74
14:46:40 <zkis> OK, I can take a look if I would have any useful suggestions on this
14:47:12 <anssik> TOPIC: Workshop proceedings at TPAC 2020
14:47:36 <anssik> anssik: WebML CG Virtual Meeting at TPAC 2020 discusses the W3C Workshop on Web and Machine Learning key outcomes with a focus on proposed near-term and long-term standardization directions and next steps.
14:47:42 <anssik> ... WebML CG participants are invited.
14:47:46 <anssik> ... meeting info:
14:47:52 <anssik> ... WebML CG Virtual Meeting at TPAC 2020
14:47:52 <anssik> ... 22 October 2020 - 14:00-15:00 UTC+0
14:47:52 <anssik> ... Agenda: https://github.com/webmachinelearning/meetings/tree/master/2020-10-22-tpac2020
14:48:29 <anssik> TOPIC: Proposals for future work
14:48:54 <anssik> anssik: Per workshop feedback, new repo created to capture Web & ML proposals for future work that are not in scope of any group currently:
14:48:59 <anssik> -> https://github.com/webmachinelearning/proposals WebML CG proposals repo
14:50:20 <anssik> TOPIC: Adjourn
14:50:23 <ping_yu> bye
14:50:28 <anssik> RRSAgent, draft minutes v2
14:50:28 <RRSAgent> I have made the request to generate https://www.w3.org/2020/10/01-webmachinelearning-minutes.html anssik
14:58:42 <anssik> Present+ Zoltan_Kis
14:58:45 <anssik> RRSAgent, draft minutes v2
14:58:45 <RRSAgent> I have made the request to generate https://www.w3.org/2020/10/01-webmachinelearning-minutes.html anssik