Update handler.py

handler.py

@@ -1,21 +1,25 @@
 # handler.py
 import torch
 from transformers import AutoTokenizer, AutoModelForCausalLM
+from peft import PeftModel
 
-MODEL_PATH = "GilbertAkham/deepseek-R1-multitask-lora"
+BASE_MODEL = "deepseek-ai/deepseek-coder-6.7b-instruct"  # or your real base
+ADAPTER_PATH = "GilbertAkham/deepseek-R1-multitask-lora"
 
 class EndpointHandler:
     def __init__(self, path=""):
-        print("Loading merged model...")
-        self.tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH, trust_remote_code=True)
-        self.model = AutoModelForCausalLM.from_pretrained(
-            MODEL_PATH,
+        print("Loading base model...")
+        self.tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, trust_remote_code=True)
+        base_model = AutoModelForCausalLM.from_pretrained(
+            BASE_MODEL,
             torch_dtype=torch.float16,
             device_map="auto",
             trust_remote_code=True
         )
+        print("Attaching LoRA adapter...")
+        self.model = PeftModel.from_pretrained(base_model, ADAPTER_PATH)
         self.model.eval()
-        print("✅ Model loaded successfully.")
+        print("✅ LoRA adapter loaded successfully.")
 
     def __call__(self, data):
         prompt = data.get("inputs", "")