Return predictor from deploy method via super

Author: siddvenk

src/sagemaker/djl_inference/model.py

@@ -32,7 +32,7 @@
 
 logger = logging.getLogger("sagemaker")
 
-
+# DJL Serving uses log4j, so we convert python logging name to log4j equivalent
 LOG_LEVEL_MAP = {
     logging.INFO: "info",
     logging.DEBUG: "debug",
@@ -200,7 +200,7 @@ def deploy(
                     f"with GPUs. Supported instance families are {defaults.ALLOWED_INSTANCE_FAMILIES}"
                 )
 
-        super(DJLLargeModel, self).deploy(
+        return super(DJLLargeModel, self).deploy(
             initial_instance_count=initial_instance_count,
             instance_type=instance_type,
             serializer=serializer,
@@ -447,8 +447,8 @@ def __init__(
         self.engine = DJLEngine.HUGGINGFACE_ACCELERATE
         self.device_id = device_id
         self.device_map = device_map
-        self.load_in_8bit = (load_in_8bit,)
-        self.low_cpu_mem_usage = (low_cpu_mem_usage,)
+        self.load_in_8bit = load_in_8bit
+        self.low_cpu_mem_usage = low_cpu_mem_usage
 
     def generate_serving_properties(self, serving_properties={}) -> Dict[str, str]:
         serving_properties = super(HuggingfaceAccelerateModel, self).generate_serving_properties()