Update controlnet/callable_functions.py

controlnet/callable_functions.py

@@ -10,22 +10,14 @@ from transformers import AutoProcessor, SiglipVisionModel
 
 
 
-def use_stylecode(model,image_path, prompt,negative_prompt, num_inference_steps, stylecode,image=None):
+def use_stylecode(model,image_path, prompt,negative_prompt, num_inference_steps, stylecode,seed=None,image=None):
     # Load and preprocess image
     # Set up model components
     unet = UNet2DConditionModel.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder="unet", torch_dtype=torch.float16, device="cuda")
     stylecodes_model = StyleCodesModel.from_unet(unet, size_ratio=1.0).to(dtype=torch.float16, device="cuda")
 
-    noise_scheduler = DDIMScheduler(
-        num_train_timesteps=1000,
-        beta_start=0.00085,
-        beta_end=0.012,
-        beta_schedule="scaled_linear",
-        clip_sample=False,
-        set_alpha_to_one=False,
-        steps_offset=1,
-    )
 
+    print("running prompt = ",prompt, " negative_prompt = ",negative_prompt, " with code ", stylecode)
     stylecodes_model.load_model(model)
 
     pipe = StableDiffusionPipelineXSv2.from_pretrained(
@@ -46,8 +38,10 @@ def use_stylecode(model,image_path, prompt,negative_prompt, num_inference_steps,
     image = image.resize((512, 512))
 
     # Set up generator with a fixed seed for reproducibility
-    seed = 238
-    generator = torch.Generator(device="cuda").manual_seed(seed)
+    if seed is not None and not -1:
+        generator = torch.Generator(device="cuda").manual_seed(seed)
+    else:
+        generator = None
 
     # Run the image through the pipeline with the specified prompt
     output_images = pipe(