1 comment

[ 4.0 ms ] story [ 15.1 ms ] thread
A 10-min video that does a good job of explaining the token processing and the probability-based "next then" text generation process of LLMs.

I highly recommend for everyone.