Update inference.py

inference.py

@@ -1,10 +1,12 @@
-import functools
 import os
 import PIL.Image
 import torch
 from huggingface_hub import login
 from transformers import PaliGemmaForConditionalGeneration, PaliGemmaProcessor
-import spaces
+import jax
+import jax.numpy as jnp
+import numpy as np
+import functools
 
 hf_token = os.getenv("HF_TOKEN")
 login(token=hf_token, add_to_git_credential=True)
@@ -16,7 +18,6 @@ class PaliGemmaModel:
         self.model = PaliGemmaForConditionalGeneration.from_pretrained(self.model_id).eval().to(self.device)
         self.processor = PaliGemmaProcessor.from_pretrained(self.model_id)
 
-    @spaces.GPU
     def infer(self, image: PIL.Image.Image, text: str, max_new_tokens: int) -> str:
         inputs = self.processor(text=text, images=image, return_tensors="pt")
         inputs = {k: v.to(self.device) for k, v in inputs.items()}  # Move inputs to the correct device
@@ -125,4 +126,3 @@ class VAEModel:
                 return x
 
         return jax.jit(Decoder().apply, backend='cpu')
-