Update on WebML CG Teleconference cadence, Model Loader API Dev Trial

Hi WebML CG participants,

TL;DR: Our next WebML CG call opportunity is 29 June.

Given the active stage of development for the Model Loader API and expected progress in prototyping around GPU support exploration, I have made a decision to skip the WebML CG call scheduled 1 June to give our key contributors time to fully focus on the implementation work ahead of them.

Meanwhile, please do check the first-ever Model Loader API Dev Trial in M102 for Chrome OS and provide feedback! Turn on these two flags to experiment:

#enable-experimental-web-platform-features
#enable-machine-learning-model-loader-web-platform-api

Honglin shared more details in a GH issue and feedback is welcome in the same issue if related to this experiment:

https://github.com/webmachinelearning/model-loader/issues/36#issuecomment-1125497616


As a reminder, feel free to share any Model Loader API spec proposals via GH issues:

https://github.com/webmachinelearning/model-loader/issues


The explainer is currently aligned with the implementation in Dev Trial:

https://github.com/webmachinelearning/model-loader/blob/main/explainer.md

The spec is expected to receive updates to bring it on par with the implementation. It is not yet fully up to date.

If you’re creating your own experiment using the Model Loader API please refer to the explainer or use the image classification demo Honglin made available as a starting point:

https://false-shy-event.glitch.me/


Thanks Honglin and the team for delivering this experiment for the community to play with!

Thanks,

-Anssi (WebML WG/CG chair)