Update app.py

app.py

@@ -53,16 +53,16 @@ def upload_to_ftp(filename):
     except Exception as e:
         print(f"FTP upload error: {e}")
 
-device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+device = torch.device("cuda:0")
 torch_dtype = torch.bfloat16
 
 checkpoint = "microsoft/Phi-3.5-mini-instruct"
 #vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
 #vae = AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16")
 #vae = AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16")
-vae = AutoencoderKL.from_pretrained("stabilityai/sdxl-vae",use_safetensors=True,safety_checker=None)
+vae = AutoencoderKL.from_pretrained("stabilityai/sdxl-vae",safety_checker=None)
 
-pipe = StableDiffusion3Pipeline.from_pretrained("ford442/stable-diffusion-3.5-medium-bf16",use_safetensors=True).to(device=torch.device("cuda:0")).to(torch.bfloat16)
+pipe = StableDiffusion3Pipeline.from_pretrained("ford442/stable-diffusion-3.5-medium-bf16").to(device=torch.device("cuda:0")).to(torch.bfloat16)
 #pipe = StableDiffusion3Pipeline.from_pretrained("ford442/stable-diffusion-3.5-medium-bf16").to(torch.device("cuda:0"))
 #pipe = StableDiffusion3Pipeline.from_pretrained("ford442/RealVis_Medium_1.0b_bf16", torch_dtype=torch.bfloat16)
 #pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai/stable-diffusion-3.5-medium", token=hftoken, torch_dtype=torch.float32, device_map='balanced')
@@ -76,7 +76,7 @@ pipe = StableDiffusion3Pipeline.from_pretrained("ford442/stable-diffusion-3.5-me
 #pipe = torch.compile(pipe)
 # pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear")
 
-refiner = StableDiffusionXLImg2ImgPipeline.from_pretrained("ford442/stable-diffusion-xl-refiner-1.0-bf16", use_safetensors=True, requires_aesthetics_score=True)
+refiner = StableDiffusionXLImg2ImgPipeline.from_pretrained("ford442/stable-diffusion-xl-refiner-1.0-bf16", requires_aesthetics_score=True)
 refiner.vae=vae
 refiner.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear", algorithm_type="sde-dpmsolver++")
 refiner.to(device=torch.device("cuda:0"))