backup

app.py

@@ -28,8 +28,8 @@ check_environment()
 login(token=os.environ["HF_TOKEN"], add_to_git_credential=True)
 
 # Load model and processor (do this outside the inference function to avoid reloading)
-base_model_path = "meta-llama/Llama-3.2-11B-Vision-Instruct"
-lora_weights_path = "taesiri/BugsBunny-LLama-3.2-11B-Vision-Base-Medium-LoRA"
+base_model_path = "taesiri/BugsBunny-LLama-3.2-11B-Vision-Instruct-Medium-FullModel"
+# lora_weights_path = "taesiri/BugsBunny-LLama-3.2-11B-Vision-Base-Medium-LoRA"
 
 processor = AutoProcessor.from_pretrained(base_model_path)
 model = MllamaForConditionalGeneration.from_pretrained(
@@ -37,9 +37,10 @@ model = MllamaForConditionalGeneration.from_pretrained(
     torch_dtype=torch.bfloat16,
     device_map="cuda",
 )
-model = PeftModel.from_pretrained(model, lora_weights_path)
 model.tie_weights()
 
+# model = PeftModel.from_pretrained(model, lora_weights_path)
+
 
 def describe_image_in_JSON(json_string):
     try: