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(compile): use absolute engineering notes link
cklxxb504d6cba46e
diff --git a/README.md b/README.mdindex 55a48f46..129b0b60 100644--- a/README.md+++ b/README.md@@ -887,7 +887,7 @@ absent; an existing value always wins. It does nothing on eager or TileLang load`compile_mode="reduce-overhead"` is a separate opt-in for CUDA graphs. With `compile=True`, itcan reduce launch overhead on repeated shapes, but each new shape needs recording and GPU memorycan grow. Laya copies outputs out of reusable graph buffers and serializes these CUDA forwards.-The default mode remains unchanged. See [the engineering notes](docs/compile-and-fast-path.md)+The default mode remains unchanged. See [the engineering notes](https://github.com/NandhaKishorM/laya/blob/main/docs/compile-and-fast-path.md)for measurements and limitations.---