Re:Show HN: Explore large language models on any computer wi...
by george_808 on Saturday June 17, 2023 @12:50PM
(#36371935)
From what I can tell, this doesn’t re-seed the random number generator, so we shouldn’t expect deterministic results. A better test would be to examine the model’s logits, or probabilities of the next token, across different ram sizes.
'You have 8 apples.'
>>> lm.set_max_ram('4gb')
4.0
>>> lm.do("If I have 7 apples then eat 5, how many apples do I have?")
'I have 2 apples left.'
It's funny how it switched from "You" to "I" when the memory was increased.