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
test(env): cite the comparison by name, not by line number
Aashish0e65f6514306
diff --git a/tests/test_env_docs.py b/tests/test_env_docs.pyindex 341b5223..8e82cce4 100644--- a/tests/test_env_docs.py+++ b/tests/test_env_docs.py@@ -14,8 +14,8 @@ And a name being written down is not the same as what it accepts being written d``LAYA_CUDA_AMP`` and ``LAYA_CPU_AMP`` are the two names in the package whose accepted values area literal tuple, and both of the places that list them are wrong: ``docs/docker.md``'s``LAYA_CPU_AMP`` row says "``bf16`` opts the CPU forward into bf16; anything else leaves it fp32"-while ``laya/agent.py:566`` compares the value against ``("bf16", "bfloat16")``, and-``_cuda_amp_dtype``'s own docstring -- the text a maintainer editing the function reads -- says+while the CPU comparison accepts ``bf16`` and ``bfloat16`` alike, and ``_cuda_amp_dtype``'s own+docstring -- the text a maintainer editing the function reads -- says"LAYA_CUDA_AMP=fp16|bf16 when set ... anything else is ignored" while the function accepts fourspellings. The error is in the direction that hides a working knob: ``LAYA_CPU_AMP=bfloat16``turns bf16 on today and the page promises it will not, so the deployment that trusts the page