scheduling_tcd

app.py

@@ -9,6 +9,7 @@ from diffusers import AutoPipelineForImage2Image
 from diffusers.utils import load_image
 import math
 from DeepCache import DeepCacheSDHelper
+from scheduling_tcd import TCDScheduler
 
 adapter_id = "latent-consistency/lcm-lora-sdv1-5"
 
@@ -30,6 +31,10 @@ def resize(value,img):
 def infer(model_id,source_img, prompt, steps, seed, Strength):
     pipe = OVStableDiffusionImg2ImgPipeline.from_pretrained(model_id, torch_dtype=torch.float16, export=True) if torch.cuda.is_available() else AutoPipelineForImage2Image.from_pretrained("stabilityai/sdxl-turbo")
     pipe = pipe.to(device)
+    pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+    tcd_lora_id = "h1t/TCD-SDXL-LoRA"
+    pipe.load_lora_weights(tcd_lora_id)
+    pipe.fuse_lora()
     start_time = time.time()
     generator = torch.Generator(device).manual_seed(seed)  
     if int(steps * Strength) < 1:
@@ -43,7 +48,7 @@ def infer(model_id,source_img, prompt, steps, seed, Strength):
     return image
 
 gr.Interface(fn=infer, inputs=[
-    gr.Text(value="Lykon/dreamshaper-8", label="Checkpoint"),
+    gr.Text(value="Lykon/dreamshaper-xl-v2-turbo", label="Checkpoint"),
     gr.Image(sources=["upload", "webcam", "clipboard"], type="filepath", label="Raw Image."), 
     gr.Textbox(label = 'Prompt Input Text. 77 Token (Keyword or Symbol) Maximum'),
     gr.Slider(1, 5, value = 2, step = 1, label = 'Number of Iterations'),