Update app.py

app.py

@@ -122,6 +122,8 @@ def process(
     for _ in tqdm(range(num_samples)):
         shape = (1, 4, height // 8, width // 8)
         x_T = torch.randn(shape, device=device, dtype=torch.float32)
+        # Create unconditional embeddings for classifier-free guidance
+        uc = model.get_learned_conditioning([""])
 
         if not tile_diffusion:
             samples = sampler.sample_ccsr(
@@ -129,9 +131,8 @@ def process(
                 positive_prompt=positive_prompt, negative_prompt=negative_prompt, x_T=x_T,
                 cfg_scale=cfg_scale,
                 color_fix_type="adain" if use_color_fix else "none",
-                # Add unconditional embeddings for classifier-free guidance
-                unconditional_guidance_scale=cfg_scale,
-                unconditional_conditioning=None  # You might need to define this based on your model
+                # Pass unconditional embeddings to the sampler
+                unconditional_conditioning=uc,
             )
         else:
             samples = sampler.sample_with_tile_ccsr(
@@ -140,9 +141,8 @@ def process(
                 positive_prompt=positive_prompt, negative_prompt=negative_prompt, x_T=x_T,
                 cfg_scale=cfg_scale,
                 color_fix_type="adain" if use_color_fix else "none",
-                # Add unconditional embeddings for classifier-free guidance
-                unconditional_guidance_scale=cfg_scale,
-                unconditional_conditioning=None  # You might need to define this based on your model
+                # Pass unconditional embeddings to the sampler
+                unconditional_conditioning=uc,
             )
 
         x_samples = samples.clamp(0, 1)