Update app.py

app.py

@@ -139,8 +139,8 @@ def load_and_prepare_model():
         
     '''
     #pipe.vae=vaeX
-    pipe.to(device=device, dtype=torch.bfloat16)
     pipe.vae=vaeX.to(device)
+    pipe.to(device=device, dtype=torch.bfloat16)
    # pipe.vae.enable_tiling()
    # pipe.vae.fuse_qkv_projections()
     #pipe.vae.to(device=device, dtype=torch.bfloat16)
@@ -167,8 +167,8 @@ pipe = load_and_prepare_model()
 # text models
 #checkpoint = "microsoft/Phi-3.5-mini-instruct"
 checkpoint = "ford442/Phi-3.5-mini-instruct-bf16"
-captioner_2 = pipeline(model="ydshieh/vit-gpt2-coco-en",device='cuda:0', task="image-to-text")
-#captioner = pipeline(model="Salesforce/blip-image-captioning-base",device='cuda', task="image-to-text")
+#captioner = pipeline(model="ydshieh/vit-gpt2-coco-en",device='cuda:0', task="image-to-text")
+captioner_2 = pipeline(model="Salesforce/blip-image-captioning-base",device='cuda', task="image-to-text")
 #captioner_3 = pipeline(model="ford442/blip-image-to-text-large-bf16",device='cuda', task="image-to-text")
 model5 = Blip2ForConditionalGeneration.from_pretrained("ford442/blip2-image-to-text-bf16").to('cuda')
 processor5 = Blip2Processor.from_pretrained("ford442/blip2-image-to-text-bf16", device_map='cuda')