Update app.py

app.py

@@ -106,32 +106,33 @@ def apply_style(style_name: str, positive: str, negative: str = "") -> Tuple[str
 def load_and_prepare_model(model_id):
     model_dtypes = {"ford442/RealVisXL_V5.0_BF16": torch.bfloat16,}
     dtype = model_dtypes.get(model_id, torch.bfloat16)  # Default to float32 if not found
-    #vae = AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16", torch_dtype=torch.bfloat16,safety_checker=None).to('cuda')
-    vae = AutoencoderKL.from_pretrained("stabilityai/sdxl-vae",use_safetensors=True, torch_dtype=torch.float32,safety_checker=None)
+    vae = AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16", torch_dtype=torch.bfloat16,safety_checker=None).to(device)
+    #vae = AutoencoderKL.from_pretrained("stabilityai/sdxl-vae",use_safetensors=True, torch_dtype=torch.float32,safety_checker=None).to(device).to(torch.bfloat16)
   #  vae = AutoencoderKL.from_pretrained("BeastHF/MyBack_SDXL_Juggernaut_XL_VAE/MyBack_SDXL_Juggernaut_XL_VAE_V10(version_X).safetensors",safety_checker=None).to(torch.bfloat16)
     #  vae = AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16", safety_checker=None).to('cuda')
     pipeX = StableDiffusionXLPipeline.from_pretrained("SG161222/RealVisXL_V5.0",use_safetensors=True, torch_dtype=torch.float32)
+    pipeX.unet.to(device).to(torch.bfloat16)
     pipe = StableDiffusionXLPipeline.from_pretrained(
         model_id,
-       # torch_dtype=torch.bfloat16,
+        torch_dtype=torch.bfloat16,
         add_watermarker=False,
         use_safetensors=True,
      #  vae=AutoencoderKL.from_pretrained("BeastHF/MyBack_SDXL_Juggernaut_XL_VAE/MyBack_SDXL_Juggernaut_XL_VAE_V10(version_X).safetensors",repo_type='model',safety_checker=None),
      #   vae=AutoencoderKL.from_pretrained("stabilityai/sdxl-vae",repo_type='model',safety_checker=None, torch_dtype=torch.float32),
        # vae=AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16",repo_type='model',safety_checker=None),
-        vae=vae,
-        unet=pipeX.unet,
-        scheduler = EulerAncestralDiscreteScheduler.from_config(pipeX.scheduler.config, beta_schedule="scaled_linear", beta_start=0.00085, beta_end=0.012, steps_offset=1)
+        #vae=vae,
+       # unet=pipeX.unet,
+      #  scheduler = EulerAncestralDiscreteScheduler.from_config(pipeX.scheduler.config, beta_schedule="scaled_linear", beta_start=0.00085, beta_end=0.012, steps_offset=1)
         #scheduler=EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear", beta_start=0.00085, beta_end=0.012, steps_offset =1)
     )
-    #pipe.vae=vae
-    #pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear", beta_start=0.00085, beta_end=0.012, steps_offset=1)
+    pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear", beta_start=0.00085, beta_end=0.012, steps_offset=1)
     #pipe.to('cuda')
    # pipe.scheduler=EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
-   # pipe.unet = pipeX.unet
     #pipe.to(dtype=torch.bfloat16)
     pipe.to(device)
     pipe.to(torch.bfloat16)
+    pipe.vae=vae
+    pipe.unet = pipeX.unet
     #pipe.to(device, torch.bfloat16)
     del pipeX
     #sched = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear", algorithm_type="dpmsolver++")