Contact emails

[email protected], [email protected], [email protected]

Explainer

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

Specification

https://webmachinelearning.github.io/model-loader/#api

Summary

Run machine learning models natively, taking advantage of hardware
acceleration. Currently, javascript libraries must parse a model and call
WebGL, WebGPU and WASM for the individual compute operations. Providing a
model loader API gains performance by optimizing in the browser and
potentially taking advantage of new ML processors like TPUs.


Blink component

Blink>WebML
<https://bugs.chromium.org/p/chromium/issues/list?q=component:Blink%3EWebML>

Motivation

Model Loader is a proposed web API to load a custom, pre-trained machine
learning model in a standard format, compile it for the available hardware,
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.

Performing inference locally can:

- Preserve privacy, by not shipping user data across the network

- Improve performance, by eliminating network latency, running models
natively and taking advantage of hardware acceleration, including
specialized hardware not available with WebGL, WebGPU or WASM.

- Provide a fallback if network access is unavailable, possibly using a
smaller and lower quality model

The graph-based Web NN API and the Model Loader API are complementary
approaches. We'll need to do some benchmarking to understand if there are
performance differences, and get feedback from developers to see if it's
valuable to offer both types of API.

Initial public proposal

https://webmachinelearning.github.io/model-loader-intro/

TAG review

TAG review status

Pending

Risks

The Model Loader API is currently included as a Tentative Specification in
the Web Machine Learning Working Group Charter.


Interoperability and Compatibility

Gecko: No signals

WebKit: No signals


Chrome/TensorFlow: Positive

Intel : Collaborating in Working Group

Microsoft/ONNX: Collaborating in Working Group

PyTorch: No signals

Salesforce: Collaborating in Working Group



Debuggability

Is this feature fully tested by web-platform-tests
<https://chromium.googlesource.com/chromium/src/+/master/docs/testing/web_platform_tests.md>
?

No

Flag name

Requires code in //chrome?

False

Tracking bug

https://bugs.chromium.org/p/chromium/issues/detail?id=1263240

Estimated milestones

No milestones specified


Link to entry on the Chrome Platform Status

https://www.chromestatus.com/feature/6505078216196096

This intent message was generated by Chrome Platform Status
<https://www.chromestatus.com/>.

-- 
You received this message because you are subscribed to the Google Groups 
"blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAEK6eFxSmK6ajiQ6_7rhPspiaDXO3xM3uNJnbSgCobQ_nk%2Bxpg%40mail.gmail.com.

Reply via email to