Ambix
Joined 18 years ago
Karma
60
Submissions
103
Member since
Dec 4
Joined 18 years ago
For bigger models (in range of 8B - 70B) the Q4KM is very good, there are no any degradation compared to full FP16 models.
Which table format is better for LLMs? Do you have some insights there?
Recently I have been busy writing the emulator in Golang:
But I believe that most of the data stored in foundation models are just useless for some particular domain. So it's better to forget something, getting really useful info instead.
For example, I use only 6 cores from 10 on my M1 Pro laptop.