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(evals): a JSON container subclass encodes as the mapping it wraps
Aashishdb74aae23067
diff --git a/laya/_agreement.py b/laya/_agreement.pyindex 3eb4c506..c46ab2c9 100644--- a/laya/_agreement.py+++ b/laya/_agreement.py@@ -24,14 +24,23 @@ def _text(value: Any, where: str) -> None:def _json(value: Any, where: str) -> str:- """Strict JSON, preserving object order because it can change model inputs."""+ """Strict JSON, preserving object order because it can change model inputs.++ Containers are matched by `isinstance`, scalars by exact type. `json.dumps` writes a `dict` or+ `list` subclass as the very object or array it wraps -- an `OrderedDict`, a `defaultdict`, a+ `Counter`, or anything read back with `object_pairs_hook=OrderedDict`, the way a caller keeps the+ key order this module fingerprints -- encodes byte-for-byte as the mapping it equals and hashes to+ the same fingerprint, so matching the encoder here refuses nothing the encoder would have taken.+ Scalars are matched by exact type: `numpy.float64` and `numpy.str_` do encode, but as a type the+ writer did not declare, and `numpy.int64` and `numpy.bool_` raise `TypeError` in the encoder.+ """def check(item):- if type(item) is dict:+ if isinstance(item, dict):for key, child in item.items():if not isinstance(key, str):raise ValueError("object keys must be strings")check(child)- elif type(item) is list:+ elif isinstance(item, list):for child in item:check(child)elif type(item) not in (str, int, float, bool, type(None)):