Update app.py

app.py

@@ -116,10 +116,8 @@ def load_and_prepare_model(model_id):
         add_watermarker=False,
         use_safetensors=True,
         vae=vae,
-    )
+    ).to('cuda')
     pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
-    pipe.to(torch.bfloat16)
-    pipe.to('cuda')
     if ENABLE_CPU_OFFLOAD:
         pipe.enable_model_cpu_offload()
     return pipe
@@ -175,7 +173,7 @@ def generate(
     seed: int = 1,
     width: int = 768,
     height: int = 768,
-    guidance_scale: float = 4.2,
+    guidance_scale: float = 3.0,
     num_inference_steps: int = 250,
     randomize_seed: bool = False,
     use_resolution_binning: bool = True, 
@@ -226,7 +224,7 @@ def generate_cpu(
     seed: int = 1,
     width: int = 768,
     height: int = 768,
-    guidance_scale: float = 5,
+    guidance_scale: float = 3,
     num_inference_steps: int = 250,
     randomize_seed: bool = False,
     use_resolution_binning: bool = True, 
@@ -382,7 +380,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
                 minimum=0.1,
                 maximum=6,
                 step=0.1,
-                value=4.2,
+                value=3.0,
             )
             num_inference_steps = gr.Slider(
                 label="Number of inference steps",