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(code): remove a duplicated docstring line and two stale paths
Ash60e5a60fba35
diff --git a/assets/make_logo.py b/assets/make_logo.pyindex 839e7b1c..3d73e72f 100644--- a/assets/make_logo.py+++ b/assets/make_logo.py@@ -9,7 +9,7 @@ neither one the end. It carries a second meaning specific to this project -- a cresolving into discrete points is what the model does, turning unstructured state into a typeddecision.- python3 notebooks/make_logo.py -> assets/logo-mark.svg logo-mark.png+ python3 assets/make_logo.py -> assets/logo-mark.svg logo-mark.pngassets/logo-lockup.svg logo-lockup.pngassets/logo-mark-mono.svg"""diff --git a/examples/_common.py b/examples/_common.pyindex d35333d8..48cb112b 100644--- a/examples/_common.py+++ b/examples/_common.py@@ -2,7 +2,7 @@Each example is a standalone script:- .venv/bin/python examples/01_hello_world_decision.py+ .venv/bin/python examples/02_hello_world_decision.pyRun it from anywhere -- paths resolve from this file, and Python puts the examples/ directoryon `sys.path` automatically, so `from _common import ...` just works.diff --git a/laya/cli.py b/laya/cli.pyindex 4b46c774..94616f48 100644--- a/laya/cli.py+++ b/laya/cli.py@@ -19,8 +19,7 @@ which needs network access to the Hugging Face hub. --preset answers one of theready-made question presets from laya.presets and implies --predict.--batch scores a whole file (or stdin, with `-`) through Router.predict_batch in oneprocess, so the states share checkpoint loads and forward passes; --json prints JSONL.-ready-made question presets from laya.presets and implies --predict. --questions-answers the question set in a JSON file instead; --max-len and --head-max-len+--questions answers the question set in a JSON file instead; --max-len and --head-max-lenraise its token budget for the request, which is what a many-label question needs,and --min-confidence marks an answer whose confidence falls below the thresholdinstead of changing it (it needs an answering mode: routing has no confidence to gate).diff --git a/laya/router.py b/laya/router.pyindex 33bb0aee..f1dc67df 100644--- a/laya/router.py+++ b/laya/router.py@@ -466,7 +466,7 @@ class Router(HookRegistry):against, read it back: the map handed to each `Agent` is the merge described above.Both `agent_kwargs` and `sha256_digests` are public and mutable, and a checkpoint's entry is read- on the load rather than at construction, so a pin assigned in afterwards -- or added to an+ on the load rather than at construction, so a pin assigned afterwards -- or added to anexisting entry in place -- counts.The names the Router sets for itself -- `model_id_or_path`, `device`, `token`, `subfolder`,