Open-Assistant
Open-Assistant copied to clipboard
LAION-AI
OpenAssistant is a chat-based assistant that understands tasks, can interact with third-party systems, and retrieve information dynamically to do so.
Changed the regex in utils.py - the original was missing references with 2+ ids and spaces after commas. Also changed the way the WebGPT is loaded to use dataset.map() instead...
There is currently no way to visualize the data, but a script or command can be added if necessary.
This also broadcasts task completion messages to every server. This needs to be changed in the future if the bot starts sending too many messages, but it should be fine...
The bot now keeps a `set` of user ID's for everyone currently doing a task. This gets removed after a user times out or doesn't want to do another task.
As we're developing more features that might not be complete-able within a single PR, we might benefit from adopting some kind of feature flag system where by the feature requires...
Humans learn how to align their actions with the goals of others through feelings of pride and shame. As a human, If someone thanks me for doing a good job,...
We can deploy the open assistant website to AWS for scalability but we can't yet deploy the backend. This requires: - [x] Support for a backend specific PostgresDB - [...
Use the buttons in CallToAction.tsx as a base. Create Discord and Github buttons which can take arbitrary text like "Continue with Discord" / "Sign in with Discord" / "Join Our...
Creating augmented data using few-shot prompts for explanations of jokes, logical inferences, etc.
See https://www.lesswrong.com/posts/EHbJ69JDs4suovpLw/testing-palm-prompts-on-gpt3. Try doing 2, 3 or 4 shot inference on something like JT or neox 20B or galactica. After we find a promising model and configuration, we can scrape...
