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
fix(train): drop a function-local warnings import that shadowed the module
Nandakishor6ab95ea63467
diff --git a/laya/train.py b/laya/train.pyindex 076b574b..cca4e6a0 100644--- a/laya/train.py+++ b/laya/train.py@@ -816,8 +816,11 @@ def finetune(data: str, model_dir: str, output_dir: str, config: Optional[TrainCif qid not in sample_questions:sample_questions[qid] = qdefelif sample_questions[qid] != qdef:- import warnings- warnings.warn("Conflicting schema detected for question %r across rows; keeping first seen definition." % (qid,))+ # `warnings` is imported at module scope. A function-local `import warnings`+ # here would make the name local for the whole of `finetune`, and the+ # calibration warnings above it would then raise UnboundLocalError.+ warnings.warn("Conflicting schema detected for question %r across rows; "+ "keeping first seen definition." % (qid,))if sample_questions:for d in (output_dir, os.path.join(output_dir, "checkpoint_latest")):os.makedirs(d, exist_ok=True)