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
Merge pull request #810 from Swatantra-66/fix/concurrent-gpu-oom-fallback
Nandakishore54b34002b00
diff --cc tests/test_agent_gate.pyindex 33f862fe,d9c014c2..09d69fe0--- a/tests/test_agent_gate.py+++ b/tests/test_agent_gate.py@@@ -16,14 -16,13 +16,16 @@@ import typeROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))sys.path.insert(0, ROOT)- if "torch" not in sys.modules:+ try:+ import torch+ except ImportError:t = types.ModuleType("torch")- t.cuda = types.ModuleType("torch.cuda")- t.cuda.OutOfMemoryError = type("OutOfMemoryError", (RuntimeError,), {})+ # laya/common.py annotates `proper_reward` with both at module scope, so importing+ # laya.agent through this stub needs them present even though no test touches a tensor.t.Tensor = type("Tensor", (), {})+ t.dtype = type("dtype", (), {})+ t.cuda = types.ModuleType("torch.cuda")+ t.cuda.OutOfMemoryError = type("OutOfMemoryError", (RuntimeError,), {})t.nn = types.ModuleType("torch.nn")t.nn.Module = type("Module", (), {})t.nn.MultiheadAttention = type("MultiheadAttention", (), {})