Download
Backtrack Sampler is a framework designed for experimenting with custom sampling strategies for language models (LLMs), enabling the ability to rewind and revise generated tokens. It allows developers to create and test their own token generation strategies by providing a base structure for manipulating logits and probabilities, making it a flexible tool for those interested in fine-tuning the behavior of LLMs.
Features
- Customizable token generation strategies
- Ability to backtrack and revise generated tokens
- Integration with models from Transformers and Llama.cpp
- Anti-slop strategy to prevent undesirable token generation
- Creative writing strategy to enhance model creativity by modifying token selection
- Easy-to-implement strategies using the base strategy class
Project Samples
Categories
FrameworksLicense
MIT LicenseFollow Backtrack Sampler
Other Useful Business Software
Simplify IT and security with a single endpoint management platform
NinjaOne automates the hardest parts of IT, delivering visibility, security, and control over all endpoints for more than 20,000 customers.
The NinjaOne automated endpoint management platform is proven to increase productivity, reduce security risk, and lower costs for IT teams and managed service providers. The company seamlessly integrates with a wide range of IT and security technologies. NinjaOne is obsessed with customer success and provides free and unlimited onboarding, training, and support.
Rate This Project
Login To Rate This Project
User Reviews
Be the first to post a review of Backtrack Sampler!