News for hackers. Stuff that matters.

A comment on Show HN: Explore large language models on any computer with 512MB of RAM.

Re:Show HN: Explore large language models on any computer wi... by GaggiX on Saturday June 17, 2023 @12:03PM (#36371424)
In case anyone is wondering, this repo uses various fine-tuned Flan-T5 models.
[ Parent | on HN ]
Re:Show HN: Explore large language models on any computer wi... by rolisz on Saturday June 17, 2023 @12:32PM (#36371732)
And Flan-T5 models are awesome! Fastchat is a finetuned 3b parameter Flan-T5 model that can perform really well, comparably to llama 13b
[ Parent | on HN ]
Re:Show HN: Explore large language models on any computer wi... by jncraton on Saturday June 17, 2023 @12:44PM (#36371874)
That's correct. The current base model is an int8 quantization of LaMini-Flan-T5-248M described here:

https://github.com/mbzuai-nlp/lamini-lm

I shared more details over on Reddit:

https://www.reddit.com/r/LocalLLaMA/comments/14btk3a/explore...

[ Parent | on HN ]