Rtx 3090 - Very interesting! Falcon 40B makes Meta's LLaMA pretty irrelevant and it should be able to run on an RTX 3090 or a 4090 with 4-bit quantization (close call). This means you can have a free, local, uncensored ChatGPT-3.5-quality model because fine-tunes should pop up soon. https://t.co/R05RsmAS3t

1:49pm PDT - May 31st, 2023

Very interesting! Falcon 40B makes Meta's LLaMA pretty irrelevant and it should be able to run on an RTX 3090 or a 4090 with 4-bit quantization (close call). This means you can have a free, local, uncensored ChatGPT-3.5-quality model because fine-tunes should pop up soon. https://t.co/R05RsmAS3t

2 retweets 0 likes -

Lech Mazur (@LechMazur)