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(research): don't echo exception text in the capacity bench server's 400
Bhushan Kinge61f42e8e4d31
diff --git a/research/scripts/bench_nvidia_capacity.py b/research/scripts/bench_nvidia_capacity.pyindex 9a23a82c..49ad402a 100644--- a/research/scripts/bench_nvidia_capacity.py+++ b/research/scripts/bench_nvidia_capacity.py@@ -189,8 +189,8 @@ def serve(args):await q.put(((body.get("state", ""), body["questions"]), fut))try:return web.json_response(await fut)- except ValueError as e:- raise web.HTTPBadRequest(text=str(e))+ except ValueError: # a question the checkpoint cannot encode; the reason stays server-side+ raise web.HTTPBadRequest(text="question cannot be encoded (options exceed head_max_len?)") from Noneasync def healthz(_):return web.json_response({"ok": True})