Skip to content

armgit / dspy

Public Imported from GitHub
0

DSPy: The framework for programming—not prompting—language models

Python

Update README.md

Omar Khattab76302ae91ea6
diff --git a/README.md b/README.md
index 6a9b28358..f794c2f63 100644
--- a/README.md
+++ b/README.md
@@ -17,7 +17,7 @@
**DSPy is a framework for algorithmically optimizing the quality of LM pipelines**, which are programs that call LMs one or more times to solve a task. To use LMs to build a complex system without DSPy, you generally have to: (1) break the problem down into steps, (2) prompt your LM well until each step works well in isolation, (3) tweak the steps to work well together now, (4) generate synthetic data to tune each step, and (5) use that data to finetune smaller LMs to cut costs.
-Composing these into the right pipelines is a hard and messy process: every time you change your pipeline, your LM, or your data, all prompts may need to change. To make this more systematic, **DSPy** does two things. First, it separates the flow of your program (`modules`) from the parameters (LM prompts and weights) of each step. Second, **DSPy** introduces new `optimizers`, which are LM-driven algorithms that can craft and tune the parameters of your pipeline, given a `metric` you want to maximize.
+The real expressive power with LMs is in composition, but composing these operations into the right pipelines is a hard and messy process: every time you change your pipeline, your LM, or your data, all prompts (or finetuning steps) may need to change. To make this more systematic, **DSPy** does two things. First, it separates the flow of your program (`modules`) from the parameters (LM prompts and weights) of each step. Second, **DSPy** introduces new `optimizers`, which are LM-driven algorithms that can craft and tune the parameters (prompts and weights) of your pipeline, given a `metric` you want to maximize.
**DSPy** can routinely teach powerful models like `GPT-3.5` or `GPT-4` and local models like `T5-base` or `Llama2-13b` to be much more reliable at tasks, i.e. having higher quality and/or avoiding specific failure patterns. **DSPy** optimizers will "compile" the _same_ program into _different_ instructions, few-shot prompts, and/or weight updates (finetunes) for each LM. This is an entirely new paradigm and level of abstraction, wher LMs, their weights, and their prompts fade into the background as optimizable pieces of a larger system that can learn from data in new ways. **tldr;** less prompting, higher scores, and a more systematic approach to solving tasks with LMs.