Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

atkgen: Establish pipeline toolkit from target (represented as a detector) to trained model #980

Open
leondz opened this issue Nov 11, 2024 · 0 comments
Labels
architecture Architectural upgrades probes Content & activity of LLM probes

Comments

@leondz
Copy link
Collaborator

leondz commented Nov 11, 2024

Build atkgen toolkit

  • has access to datasets of LLM conversations
  • given a function that detects a behaviour, identify conversations where the llm exhibits this behaviour
  • identify to formats of training data:
    • continuation-based, in target turn : atkgen turn format
    • conversation-based, with whole threads
  • can train on a variety of models. target set:
    • llama
    • gpt2
    • small nemotron
    • rwkv
@leondz leondz added architecture Architectural upgrades probes Content & activity of LLM probes labels Nov 11, 2024
@leondz leondz added this to the 24.12 milestone Nov 11, 2024
@leondz leondz removed this from the 24.12 milestone Nov 25, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
architecture Architectural upgrades probes Content & activity of LLM probes
Projects
None yet
Development

No branches or pull requests

1 participant