News for hackers. Stuff that matters.

A comment on CarperAI announces plans for the first open-source “instruction-tuned” LM.

Re:CarperAI announces plans for the first open-source “instr... by lee101 on Thursday October 20, 2022 @04:37AM (#33272157)
for now I'd checkout T0 or bloom, lots of models can be instructed already.

I created https://text-generator.io and have a whole bunch of examples on there of instruction following style prompts

[ Parent | on HN ]
Re:CarperAI announces plans for the first open-source “instr... by espadrine on Thursday October 20, 2022 @06:06AM (#33272603)
How do you instruct BLOOM? Does it need fine-tuning?
[ Parent | on HN ]
Re:CarperAI announces plans for the first open-source “instr... by mritchie712 on Thursday October 20, 2022 @06:37AM (#33272777)
have you considered SQL generation with two inputs:

1. data from the information_schema of the database

2. a natural language question

I've played around with gpt3 but not providing a full schema for the database is a glaring issue with accuracy of the SQL generated.

[ Parent | on HN ]
Re:CarperAI announces plans for the first open-source “instr... by petersonh on Thursday October 20, 2022 @09:36AM (#33274494)
I've done something similar with gpt3 (codex) and had good results
[ Parent | on HN ]
Re:CarperAI announces plans for the first open-source “instr... by machiaweliczny on Thursday October 20, 2022 @07:42AM (#33273191)
Has anyone tested OTP 170B from Meta? Seems like it's public
[ Parent | on HN ]
Re:CarperAI announces plans for the first open-source “instr... by monkmartinez on Thursday October 20, 2022 @01:30PM (#33277953)
Just to get OTP up for inference would require a very large spend. To use GPT-NeoX (20B parameters) for inference requires 45GB of vRAM minimally. Its hard for me to imagine using a 170B model for fun somewhere unless one has a large GPU farm or lots of money.
[ Parent | on HN ]
Re:CarperAI announces plans for the first open-source “instr... by stellaathena on Thursday October 20, 2022 @05:05PM (#33280682)
GPT-NeoX-20B was specifically targeted to fit on A40s, A6000s, and a pair of 3090 Tis. Anything larger than that is going to be a real struggle for people who don’t own computing clusters to use.
[ Parent | on HN ]