Making some progress on training OSS alternatives to #ChatGPT on #NuNet.

27 Feb 2023, 13:49
Making some progress on training OSS alternatives to #ChatGPT on #NuNet This is a service ran on NuNets #Decentralized GPU ML use case, funded by #Cardano #ProjectCatalyst Initially, the ChatBot responded with basic 2-5 word answers & now we are getting paragraphs back 🤝

Same news in other sources

1
NuNet
NuNetNTX #766
Telegram
27 Feb 2023, 13:51
Making some progress on training OSS alternatives to ChatGPT on NuNet This is a service ran on NuNet's Decentralized GPU ML use case, funded by Cardano Project Catalyst Initially, the ChatBot responded with basic 2-5 word answers & now we are getting paragraphs back 🤝 Although the information may not be 100% accurate it is a step in the right direction ⭐️ Really fascinating to see daily improvements The model it currently is using is Thanks to our amazing Systems Scientist Avi who has been working on this 💙 You can try out the bot here: The interface allows hosting for 3 days maximum, but we can fire up a new instance whenever we want to Currently, it can take up to 512 words approximately as input & generates a maximum of 1024 as output
Making some progress on training OSS alternatives to ChatGPT on NuNet.
Making some progress on training OSS alternatives to ChatGPT on NuNet This is a service ran on NuNet's Decentralized GPU ML use case, funded by Cardano Project Catalyst Initially, the ChatBot responded with basic 2-5 word answers & now we are getting paragraphs back 🤝 Although the information may not be 100% accurate it is a step in the right direction ⭐️ Really fascinating to see daily improvements The model it currently is using is https://huggingface.co/google/flan-t5-large Thanks to our amazing Systems Scientist Avi who has been working on this 💙 You can try out the bot here: https://d14e48da035c1d1e2f.gradio.live The interface allows hosting for 3 days maximum, but we can fire up a new instance whenever we want to Currently, it can take up to 512 words approximately as input & generates a maximum of 1024 as output