Update handler.py

handler.py

@@ -4,9 +4,11 @@ import json
 
 class EndpointHandler():
   def __init__(self, path="ceiteach/chart-no-pretrain-llama31-unsloth"):
+    base_model_name = "unsloth/Meta-Llama-3.1-8B-bnb-4bit"
     model_name = "ceiteach/chart-no-pretrain-llama31-unsloth"
+    model = AutoModelForCausalLM.from_pretrained(base_model_name)
+    model.load_adapter(model_name)
     tokenizer = AutoTokenizer.from_pretrained(model_name)
-    model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype = torch.float16, device_map="auto")
     self.model = model
     self.tokenizer = tokenizer