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(readme): price the option budget as a ratio, not an absolute latency
Aashish39769db39740
diff --git a/README.md b/README.mdindex 78452a61..f37cb797 100644--- a/README.md+++ b/README.md@@ -210,8 +210,8 @@ Routing alone never downloads a checkpoint, so it returns in milliseconds. `--pr`request`. It implies `--predict`, and a question set with many labels usually wants`--head-max-len` with it: on 58 MASSIVE-INTENT labels written as one choice question, the Englishcheckpoint goes from 24/58 correct at its default 192-token option budget to 34/58 at-`--head-max-len 384` (163 ms to 235 ms per request, CPU). Widening it further costs the accuracy-back, because `max_len` then leaves fewer tokens for the request itself. [Honest+`--head-max-len 384`, for about 1.4x the per-request time on CPU. Widening it further costs the+accuracy back, because `max_len` then leaves fewer tokens for the request itself. [Honestlimits](#honest-limits) describes the same budget ceiling for a 77-option question.---