You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Platforms (browsers, operating systems, local model servers) will also provide models that it would be useful to access from web applications. How can we do this in a way consistent with downloadable models, so that if the model exists on the platform, the download can be avoided?
One suggestion: Make such models "appear" in the graph (full model) cache and usable from the WebNN API as "pre-built" models.
Platforms (browsers, operating systems, local model servers) will also provide models that it would be useful to access from web applications. How can we do this in a way consistent with downloadable models, so that if the model exists on the platform, the download can be avoided?
One suggestion: Make such models "appear" in the graph (full model) cache and usable from the WebNN API as "pre-built" models.
Related issue: webmachinelearning/webnn#1
The text was updated successfully, but these errors were encountered: