Update app.py

app.py

@@ -15,10 +15,11 @@ lora_repo = "alinasdkey/unsloth-pret-lora"
 #  Load base model
 model, processor = FastVisionModel.from_pretrained(
     model_name = model_name,
-    device_map = "cpu",
-    torch_dtype = torch.float32,
-    load_in_4bit =True,
-    load_in_8bit = False,
+    device_map = "auto",                 # or "cuda" if you're sure you're on GPU
+    load_in_4bit = False,                # Use 8bit if you're on CPU
+    load_in_8bit = True,                 # Set this to True
+    # remove torch_dtype entirely
+)
 )
 
 #  Applying LoRA adapter from the root of the repo)