DSPy: The framework for programming—not prompting—language models
Python
Fix Retry for TypedPredictors
Alex7f9ab7be95fc
diff --git a/dspy/predict/retry.py b/dspy/predict/retry.pyindex dcef1488f..f851addd3 100644--- a/dspy/predict/retry.py+++ b/dspy/predict/retry.py@@ -32,6 +32,12 @@ class Retry(Predict):return signaturedef forward(self, *, past_outputs, **kwargs):+ # Take into account the possible new signature, as in TypedPredictor+ new_signature = kwargs.pop("new_signature", None)+ if new_signature:+ self.original_signature = new_signature+ self.new_signature = self._create_new_signature(self.original_signature)+# Convert the dict past_outputs={"answer": ...} to kwargs# {past_answer=..., ...}for key, value in past_outputs.items():@@ -42,7 +48,7 @@ class Retry(Predict):# Note: This only works if the wrapped module is a Predict or ChainOfThought.kwargs["new_signature"] = self.new_signaturereturn self.original_forward(**kwargs)-+def __call__(self, **kwargs):cached_kwargs = copy.deepcopy(kwargs)kwargs["_trace"] = False@@ -59,7 +65,7 @@ class Retry(Predict):# now pop multiple reserved keys# NOTE(shangyin) past_outputs seems not useful to include in demos,# therefore dropped- for key in ["_trace", "demos", "signature", "config", "lm", "past_outputs"]:+ for key in ["_trace", "demos", "signature", "new_signature", "config", "lm", "past_outputs"]:kwargs.pop(key, None)if dsp.settings.trace is not None: