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
Commits on main
test(evals): pin that the bootstrap's memory batching cannot move the interval
b3bad0f Aashish committed 1 month ago
test(email): pin the max_chars budget's slice semantics the port fixtures record
3db3020 Aashish committed 1 month ago
test(evals): assert the directory refusal by leaf name, not the runner's temp path
808288b Aashish committed 1 month ago
perf(agent): reuse sequence stats option count instead of re-rendering options
0e07dc4 Hamidreza committed 1 month ago
fix(serve): read a blank LAYA_PORT as unset like the rest of the family
677b063 Aashish committed 1 month ago
fix(calibrate): read bins and min_bucket_n as the module's own positive integers
4f76fb6 Aashish committed 1 month ago
ci(go): generate predict_long goldens and raise the test floor to 1409
4990a97 Adeshina Lasisi committed 1 month ago
test(go): PredictLong unit, golden and real-graph tests
550f8cd Adeshina Lasisi committed 1 month ago
fix(serve): refuse a bad LAYA_LOG_LEVEL with a message, not a traceback
9628771 Aashish committed 1 month ago
test(confidence): make the finite check and the float() coercion each observable
17daa14 Aashish committed 1 month ago
fix(confidence): refuse a negative probability instead of a Jev confidence outside [0, 1]
12dca7e Aashish committed 1 month ago
test(common): report a refused numpy int bins count instead of aborting the suite
0889ac5 Aashish committed 1 month ago