quant now select exllama kernel (#1364)

Qubitium · web-flow · commit cb786e57710c · 2025-03-02T10:49:33.000+08:00
Signed-off-by: Qubitium &lt;Qubitium@modelcloud.ai&gt;
diff --git a/tests/models/test_opt.py b/tests/models/test_opt.py
@@ -24,7 +24,7 @@ class TestOpt(ModelTest):
     NATIVE_ARC_CHALLENGE_ACC = 0.1894
     NATIVE_ARC_CHALLENGE_ACC_NORM = 0.2278
 
-    KERNEL_QUANT = {AUTO_SELECT_BACKEND_ORDER[BACKEND.TRITON]}
+    KERNEL_QUANT = {AUTO_SELECT_BACKEND_ORDER[BACKEND.EXLLAMA_V1]}
     KERNEL_INFERENCE = {AUTO_SELECT_BACKEND_ORDER[BACKEND.MARLIN]}
 
     def test_opt(self):
