News for hackers. Stuff that matters.

A comment on Ask HN: When/where can we get an offline GPT3 type chatbot?.

Re:Ask HN: When/where can we get an offline GPT3 type chatbot? by tripplyons on Wednesday December 07, 2022 @06:23PM (#33901834)
I think the best publicly available model that can follow instructions right now is https://huggingface.co/bigscience/bloomz.

It has 176 billion trainable parameters, but I think it uses up terabytes of memory, so there is a trade off between model size and the ability of the model.

On most GPUs, you should be able to run https://huggingface.co/google/flan-t5-large. It is pretty good and is trained to follow instructions.

[ Parent | on HN ]
Re:Ask HN: When/where can we get an offline GPT3 type chatbot? by piecerough on Wednesday December 07, 2022 @07:00PM (#33902179)
InstructGPT is actually 1.3B, you don't need 175B like initial GPT-3 model: https://openai.com/blog/instruction-following/

ChatGPT is likely a much better variant, but also must be still small and will likely be portable.

[ Parent | on HN ]
Re:Ask HN: When/where can we get an offline GPT3 type chatbot? by tripplyons on Wednesday December 07, 2022 @08:06PM (#33902752)
ChatGPT is based on GPT3.5 which is a series of 175B models that includes code-davinci-002, text-davinci-002, and text-davinci-003. I do not think of it as portable.
[ Parent | on HN ]
Re:Ask HN: When/where can we get an offline GPT3 type chatbot? by ActorNightly on Thursday December 08, 2022 @01:13PM (#33911097)
Would it be possible to run this entire model sequentially, dumping intermediate results to SSD?
[ Parent | on HN ]
Re:Ask HN: When/where can we get an offline GPT3 type chatbot? by PartiallyTyped on Thursday December 08, 2022 @04:21AM (#33905974)
Network distillation to the rescue! Provided that you have the means to distill in the first place :D
[ Parent | on HN ]
Re:Ask HN: When/where can we get an offline GPT3 type chatbot? by moffkalast on Thursday December 08, 2022 @11:17AM (#33909594)
I've just tried out a few of the flan-t5s and they're surprisingly coherent. It likes dogs and pine trees, thinks Makerbot is the best 3D printer lol. Can even generate some code, though it's usually wrong. And it can't seem to decide if a flower pot is orange or red.

Any idea if it's possible to chain feed a conversation into this one? I've tried a few various Q and A formats but none seem to really grab the old context.

[ Parent | on HN ]