A place for developers to talk about building on Farcaster.

/dev

Chat gpt, Claude, Perplexity all are down for me 
It's time to use my mind abit🫠

Exploring new horizons, one day at a time
A simple geek

Coding with the agility of a rat, declaring war on the computer mouse. Developer of https://www.proof-of-chad.fun/

Today I was really convinced that using self-hosted llm for daily tasks was a good decision.

e/acc || Building @ /imgn || Dogecoin core dev 2013/14

oh man, there's so much out there on huggingface - llama3-70b-instruct is definitely the king right now imo, but if you need super large (and actually working) context windows, mixtral-8x7b is amazing.

There's quantized versions of most big models there, and lots of fine-tunes for specific use-cases. Check them out!

Most agree TYSM for good suggestions 
Today, I feel the difference between open and closed models :)

Time to join the local llama group my friend

Llama3-7b GGUF can be easily fit into a low end GPU, or run straight off your cpu and sram at acceptable speeds!

Or if you need gpt4 level reasoning, cloud-hosted llama3 70b is your dude :)