A consortium of 40 tech companies, including the likes of
Facebook and Google, have come together to release a set of evaluation
benchmarks for AI. By measuring AI products against these benchmarks, companies
in the field will be able to identify optimal product solutions and, according
to the consortium, MLPerf, "take confidence" that they're deploying
the right solutions.
The benchmarks, named MLPerf Inference v0.5, center around
three common machine learning tasks: image classification, object detection and
machine translation. Given the different processing abilities of different
devices, there are separate benchmarks for AI across various platforms, such as
smartphones, servers and chips.
As well as providing best practice guidance for companies in
the AI field, it's hoped the benchmarks will help kick-start further innovation
as, despite its hype, organizations have been slow to pick up the technology.
In a statement, MLPerf's general chair Peter Mattson said, "By creating
common and relevant metrics to assess new machine learning software frameworks,
hardware accelerators, and cloud and edge computing platforms in real-life
situations, these benchmarks will establish a level playing field that even the
smallest companies can use."
No comments:
Post a Comment