Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.
Python calibrationclassificationdecision-modelhuggingfacejevmodernbertmultilingualnlppythonpytorchroutingtyped-decisionszero-shot
docs(confidence): cite the symbol, not a line number, for the OverflowError twin
Nandakishor710f903059b0
diff --git a/laya/confidence.py b/laya/confidence.pyindex 4c790a02..a2b92644 100644--- a/laya/confidence.py+++ b/laya/confidence.py@@ -67,7 +67,8 @@ def jev_confidence(answer: Dict[str, Any]) -> Optional[float]:except (TypeError, ValueError, OverflowError):# `OverflowError` is the same "does not read as a number": a JSON integer literal too wide# for a double (a 400-digit probability) overflows `float()`, and an infinite score level- # overflows `int()`. `calibrate.py:780` already names it for the same reason.+ # overflows `int()`. `apply_calibration_payload` in `calibrate.py` already catches it on+ # the `version` field for the same reason (JSON's `1e999` is `inf`).return Noneif (any(isinstance(v, bool) for v in probabilities.values())or not all(math.isfinite(v) and v >= 0.0 for v in p)):