Exploring Capabilities of Long-Context LLMs

Experimenting with Long-Context LLMs
Long-context LLMs are incredibly useful and their flexibility provides a great interface for deeper exploration of capabilities.
Here are some more thoughts on long-context LLMs and how I use them.
#ai #machinelearning #engineering

[@covertassassin1885]

Thanks for the great content!

[@edpro4776]

Is there a way for LLM to re-used the context from previous query? How can I make LLM memorize the context from previous query, and the memory happens on the LLM side.

[@elvissaravia]

I think this is only supported in a few APIs. I know that Gemini-1.5-Flash has such capability (i.e, context caching). You may find this video useful: th-cam.com/video/987Pd89EDPs/w-d-xo.htmlsi=EeYoVpZ7l4pM_w1U