Update app.py

app.py

@@ -59,13 +59,13 @@ def encode_image(image_path):
     return img_embedding  
 
 #Get the projection model
-img_proj_head = _MLPVectorProjector(512, 2560, 1, 4).to("cuda")
+img_proj_head = _MLPVectorProjector(512, 2560, 1, 4).to("cpu")
 img_proj_head.load_state_dict(torch.load('projection_finetuned.pth'))
 
 #Get the fine-tuned phi-2 model
 phi2_finetuned = AutoModelForCausalLM.from_pretrained(
     "phi2_adaptor_fineTuned", trust_remote_code=True,
-    torch_dtype = torch.float32).to("cuda")
+    torch_dtype = torch.float32).to("cpu")
     
 
 def example_inference(input_text, count): #, image, img_qn, audio):