Andreas Köpf
14fa08e2e7
* add query_incomplete_rankings() * Add SQL queries for TreeManager task selection * first working version of TreeManager.next_task() * remove old generate_task(), add mandatory_labels to text_labels task * Add ConversationMessage list to Ranking tasks * add more sophisticated sql queries to find extendible trees * add TreeManager.query_extendible_parents() * fix task validation, seed data insertion (reviewed) * provide user for task selection in text-frontend * enter 'growing' state * enter 'aborted_low_grade' state * enter 'ranking' state * check tree 'growing' state upon relpy insertion * exclude user from labeling their own messages (added DEBUG_ALLOW_SELF_LABELING setting) * add DEBUG_ALLOW_SELF_LABELING to docker-compose.yaml * fix ranking submission * add query_tree_ranking_results() * add ranked_message_ids to RankingReactionPayload * fix reply_messages instead of prompt_messages * incorment 'ranking_count' of ranked replies * added logic to check_condition_for_scoring_state * changes to msg_tree_state_machine * pre-commit changes * enter 'ready_for_scoring' state * re-add HF embedding call (lost during merge) * use prepare_conversation() helper for seed-data creation * Partially add user specified task selection Co-authored-by: Daniel Hug <danielpatrickhug@gmail.com>
Open-Assistant
Table of Contents
- What is Open Assistant?
- Do you want to try it out?
- The Plan
- The Vision
- How can you help?
- I’m in! How do I contribute?
What is Open Assistant?
Open Assistant is a project meant to give everyone access to a great chat based large language model.
We believe that by doing this we will create a revolution in innovation in language. In the same way that stable-diffusion helped the world make art and images in new ways we hope Open Assistant can help improve the world by improving language itself.
Do you want to try it out?
If you are interested in taking a look at the current state of the project, you can set up an entire stack needed to run Open-Assistant, including the website, backend, and associated dependent services.
To start the demo, run this in the root directory of the repository:
docker compose up --build
Then, navigate to http://localhost:3000 (It may take some time to boot up) and
interact with the website.
Note: When logging in via email, navigate to
http://localhost:1080to get the magic email login link.
Note: If you would like to run this in a standardized development environment (a "devcontainer") using vscode locally or in a web browser using GitHub Codespaces, you can use the provided
.devcontainerfolder.
The Plan
We want to get to an initial MVP as fast as possible, by following the 3-steps outlined in the InstructGPT paper.
- Collect high-quality human generated Instruction-Fulfillment samples (prompt + response), goal >50k. We design a crowdsourced process to collect and reviewed prompts. We do not want to train on flooding/toxic/spam/junk/personal information data. We will have a leaderboard to motivate the community that shows progress and the most active users. Swag will be given to the top-contributors.
- For each of the collected prompts we will sample multiple completions. Completions of one prompt will then be shown randomly to users to rank them from best to worst. Again this should happen crowd-sourced, e.g. we need to deal with unreliable potentially malicious users. At least multiple votes by independent users have to be collected to measure the overall agreement. The gathered ranking-data will be used to train a reward model.
- Now follows the RLHF training phase based on the prompts and the reward model.
We can then take the resulting model and continue with completion sampling step 2 for a next iteration.
The Vision
We are not going to stop at replicating ChatGPT. We want to build the assistant of the future, able to not only write email and cover letters, but do meaningful work, use APIs, dynamically research information, and much more, with the ability to be personalized and extended by anyone. And we want to do this in a way that is open and accessible, which means we must not only build a great assistant, but also make it small and efficient enough to run on consumer hardware.
Slide Decks
How can you help?
All open source projects begin with people like you. Open source is the belief that if we collaborate we can together gift our knowledge and technology to the world for the benefit of humanity.
Check out our contributing guide to get started.