diff --git a/peft/interpret/interpret.py b/peft/interpret/interpret.py
index 1492c60..bb0028b 100644
--- a/peft/interpret/interpret.py
+++ b/peft/interpret/interpret.py
@@ -1,7 +1,7 @@
 import json
 
 from transformers import AutoModelForCausalLM
-from peft import get_peft_model, PromptTuningInit, PromptTuningConfig, TaskType
+from peft import PeftModel, get_peft_model, PromptTuningInit, PromptTuningConfig, TaskType
 import torch
 
 from numpy import array
@@ -10,9 +10,7 @@
 
 
 def load_model():
-    # NOTE: Currently loading untrained model, will push changes for trained model loading
     model_name_or_path = "meta-llama/Llama-2-7b-chat-hf"
-
     peft_config = PromptTuningConfig(
         task_type=TaskType.CAUSAL_LM,
         prompt_tuning_init=PromptTuningInit.TEXT,
@@ -22,7 +20,8 @@ def load_model():
         tokenizer_name_or_path=model_name_or_path,
     )
     model = AutoModelForCausalLM.from_pretrained(model_name_or_path, device_map='cpu', torch_dtype=torch.float16)
-    model = get_peft_model(model, peft_config) # add PEFT pieces to the LLM
+    #model = get_peft_model(model, peft_config) # add PEFT pieces to the LLM
+    model = PeftModel.from_pretrained(model, 'meyceoz/prompt-llama-2')
     return model