aaronrelph

3 ms·

1. 1.▲
   by aaronrelph 4y ago

   Yeah, I mention this in the post but this variant of LLaMA isn't storing any of the conversation in memory so it doesn't have context on the prior questions. You're starting fresh with each prompt. We have some ideas for how
2. 2.▲
   by aaronrelph 4y ago

   Alright, after a few hiccups, you should be seeing a noticeable improvement in response times now!
3. 3.▲
   by aaronrelph 4y ago

   fixed!
4. 4.▲
   by aaronrelph 4y ago

   And we're back! Might be little slow while traffic is high but responses are coming through.
5. 5.▲
   by aaronrelph 4y ago

   Wow sorry y'all, we didn't expect this to take off so quickly. Working on getting this scaled up now so everyone can play!
6. 6.▲
   Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA(chatllama.baseten.co)

   402 points by aaronrelph 4y ago|215 comments
7. 7.▲
   Alpaca-30B: Facebook's 30B parameter LLaMa fine-tuned on the Alpaca dataset(abuqader.substack.com)

   8 points by aaronrelph 4y ago|1 comments
8. 8.▲
   by aaronrelph 4y ago

   A guide on how Abu fine-tuned Alpaca 30B and how to use it.
9. 9.▲
   Show HN: Hall Monitor, easily monitor the apps and services you use everyday(hallmonitor.co)

   1 points by aaronrelph 9y ago|0 comments