"default" v.s. "auto" in MLDevicePreference and MLPowerPreference

[yuhonglin]

Regarding the value that MLDevicePreference and MLPowerPreference can take, WebNN uses "default" as one value (https://webmachinelearning.github.io/webnn/#api-ml) but I feel maybe "auto" is more suitable because in this case, the backend will automatically determine the running environment. "auto" carries more message than "default". What do you think?

[yuhonglin]

@huningxin

[huningxin]

@yuhonglin Thanks for brining this up. I think this is a good proposal and a good topic for WG discussion. /cc @anssiko , WDYT?

[anssiko]

@yuhonglin webmachinelearning/webnn#169 is now cross-referenced.

Also WIP PR webmachinelearning/webnn#257 discusses whether to make device selection normative for WebNN.