The thing that's really interesting about transformers is that the weights are basically fixed after training. You load the weights once onto the chip, and then you just do inference over and over again. So if you build a chip that is specifically designed to run one model or one architecture, you can get massive efficiency gains because you're not paying for the flexibility of being able to run arbitrary code.
A concentration of a few good people solving problems far in advance—and at a fraction of the cost—of other groups by applying the simplest, most straightforward methods possible to develop and produce new projects.
A concentration of a few good people solving problems far in advance—and at a fraction of the cost—of other groups by applying the simplest, most straightforward methods possible to develop and produce new projects. All it is really is the application of common sense to some pretty tough problems.
2mo ago
Underscored — save the words that stop you in your tracks.
