So, here's this thing that's supposed to be the latest hot shit:
I download a GGML q4 quantized model from https://huggingface.co/TheBloke/Platypus2-70B-GGML
First run:
Second run:
(OK I got a bit carried away for this one) |
Third run:
Seems a bit unhinged.
That said its performance seem to get back on track after talking a bit... It doesn't like chatty conversations I think. But if you ask it serious stuff it seems pretty sane.
No comments:
Post a Comment