Skip to content

armgit / dspy

Public Imported from GitHub
0

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

Python

Update GEPA version (#8628)

Lakshya A Agrawal8f773dc75e34
diff --git a/dspy/teleprompt/gepa/gepa.py b/dspy/teleprompt/gepa/gepa.py
index 5287649e5..d7082d1ad 100644
--- a/dspy/teleprompt/gepa/gepa.py
+++ b/dspy/teleprompt/gepa/gepa.py
@@ -136,6 +136,7 @@ class GEPA(Teleprompter):
"""
GEPA is an evolutionary optimizer, which uses reflection to evolve text components
of complex systems. GEPA is proposed in the paper [GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning](https://arxiv.org/abs/2507.19457).
+ The GEPA optimization engine is provided by the `gepa` package, available from [https://github.com/gepa-ai/gepa](https://github.com/gepa-ai/gepa).
GEPA captures full traces of the DSPy module's execution, identifies the parts of the trace
corresponding to a specific predictor, and reflects on the behaviour of the predictor to
@@ -175,8 +176,8 @@ class GEPA(Teleprompter):
...
```
- GEPA can also be used as a batch inference-time search strategy, by passing `valset=trainset, track_stats=True`, and using the
- `detailed_results` attribute of the optimized program (returned by `compile`) to get the Pareto frontier of the batch.
+ GEPA can also be used as a batch inference-time search strategy, by passing `valset=trainset, track_stats=True, track_best_outputs=True`, and using the
+ `detailed_results` attribute of the optimized program (returned by `compile`) to get the Pareto frontier of the batch. `optimized_program.detailed_results.best_outputs_valset` will contain the best outputs for each task in the batch.
Example:
```
diff --git a/pyproject.toml b/pyproject.toml
index 56e7d5e5a..2350bee1d 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -42,7 +42,7 @@ dependencies = [
"rich>=13.7.1",
"numpy>=1.26.0",
"xxhash>=3.5.0",
- "gepa==0.0.1"
+ "gepa==0.0.2"
]
[project.optional-dependencies]