3 ms·
It does this poorly. They compared the currently most optimized whisper implementation for M1 against an implementation that is far from the best currently avai
by lern_too_spel 3y ago
It does this poorly. They compared the currently most optimized whisper implementation for M1 against an implementation that is far from the best currently available for the Nvidia GPU. They cannot make a claim of reaching 70% speed at 25% power usage.
Edit: The article has been updated to compare against a faster implementation for the Nvidia GPU and no longer makes that claim.