Add inference.yaml for API configuration

inference.yaml

@@ -0,0 +1,3 @@
+inference:
+  accelerator: gpu
+  parameters: {}