Update app.py

app.py

@@ -17,12 +17,12 @@ else:
     st.warning("HF_TOKEN environment variable not set. Some features may be limited.")
 
 # Model and Adapter Configuration
-model_id = "google/gemma-2b-it"  # Base model
+model_id = "Prajjwalng/gemma_customer_care"  # Base model
 adapter_id = "Prajjwalng/gemma_customercare_adapters" #adapter model
 
 # Initialize model and tokenizer (load only once)
 @st.cache_resource
-def load_model(model_id, adapter_id):
+def load_model(model_id):
     base_model = AutoModelForCausalLM.from_pretrained(
         model_id,
         low_cpu_mem_usage=True,
@@ -31,9 +31,8 @@ def load_model(model_id, adapter_id):
         device_map={"": 0} if torch.cuda.is_available() else "cpu"
     )
 
-    merged_model = PeftModel.from_pretrained(base_model, adapter_id)
     tokenizer = AutoTokenizer.from_pretrained(model_id, add_eos_token=True)
-    return merged_model, tokenizer
+    return base_model
 
 merged_model, tokenizer = load_model(model_id, adapter_id)