zou-group/textgrad
An autograd engine that uses LLMs to backpropagate textual gradients for optimizing prompts and ML system components.
Not currently ranked — collecting fresh signals.
star history
TextGrad implements automatic differentiation via text by treating LLM outputs as differentiable signals. It provides a PyTorch-like API that enables defining custom loss functions and optimizing them through text feedback from LLMs. The framework treats prompts, model outputs, and system components as optimizable variables, computing textual gradients that guide the optimization process.
Frequently asked
- What is zou-group/textgrad?
- An autograd engine that uses LLMs to backpropagate textual gradients for optimizing prompts and ML system components.
- Is textgrad open source?
- Yes — zou-group/textgrad is open source, released under the MIT license.
- What language is textgrad written in?
- zou-group/textgrad is primarily written in Python.
- How popular is textgrad?
- zou-group/textgrad has 3.7k stars on GitHub.
- Where can I find textgrad?
- zou-group/textgrad is on GitHub at https://github.com/zou-group/textgrad.