Upload app.py

app.py

@@ -30,7 +30,9 @@ base_model = AutoModelForCausalLM.from_pretrained(base_model_name, torch_dtype=t
 
 
 # Load LoRA adapter
-model = PeftModel.from_pretrained(base_model, peft_model_path)
+#model = PeftModel.from_pretrained(base_model, peft_model_path)
+model = PeftModel.from_pretrained(base_model, peft_model_path, offload_dir='./offload')
+
 model = model.merge_and_unload()  # Merge LoRA with base model
 
 # Load tokenizer