Update app.py

app.py

@@ -8,6 +8,8 @@ from PIL import Image
 
 MAX_SEED = np.iinfo(np.int32).max
 MAX_IMAGE_SIZE = 1024
+IP_ADAPTER = 'h94/IP-Adapter'
+IP_ADAPTER_WEIGHT_NAME = "ip-adapter-plus_sd15.bin"
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model_default = "stable-diffusion-v1-5/stable-diffusion-v1-5"
@@ -100,19 +102,26 @@ def infer(
 ):
     generator = torch.Generator(device).manual_seed(seed)
 
-    # Генерация с IP_adapter
+    # Генерация с Ip_Adapter
     if use_ip_adapter and ip_source_image is not None and ip_adapter_image is not None:
         pipe_ip_adapter = StableDiffusionControlNetImg2ImgPipeline.from_pretrained(    
             model_default,
             controlnet=controlnet,
             torch_dtype=torch_dtype
         ).to(device)
+
+        # Добавим Ip_Adapter 
+        pipe_ip_adapter.load_ip_adapter(IP_ADAPTER, subfolder="models", weight_name=IP_ADAPTER_WEIGHT_NAME)
+        #params['ip_adapter_image'] = load_image(ip_image).convert('RGB')
+        pipe_ip_adapter.set_ip_adapter_scale(ip_adapter_strength) # Коэфф учёта влияния Ip_Adapter на итоговое изображение
+
+        
         
         # Преобразуем изображения
         ip_source_image = preprocess_image(ip_source_image, width, height)
         ip_adapter_image = preprocess_image(ip_adapter_image, width, height)
         
-        # Создаём пайплайн IP_adapter с LoRA, если он ещё не создан   ???????????????????????????????????????????????????????????????
+        # Создаём пайплайн IP_adapter с LoRA, если он ещё не создан   
         if not hasattr(pipe_ip_adapter, 'lora_loaded') or not pipe_ip_adapter.lora_loaded:
             # Загружаем LoRA для UNet
             pipe_ip_adapter.unet = PeftModel.from_pretrained(
@@ -140,8 +149,7 @@ def infer(
         ip_adapter_strength = float(ip_adapter_strength)
         #strength_ip = float(strength_ip)
         
-        # Используем IP_adapter с LoRA  ????????????????????????????????????????????????????????????????????????
-        #pipe = pipe_ip_adapter
+        # Используем IP_adapter с LoRA  
         prompt_embeds = long_prompt_encoder(prompt, pipe_ip_adapter.tokenizer, pipe_ip_adapter.text_encoder)
         negative_prompt_embeds = long_prompt_encoder(negative_prompt, pipe_ip_adapter.tokenizer, pipe_ip_adapter.text_encoder)
         prompt_embeds, negative_prompt_embeds = align_embeddings(prompt_embeds, negative_prompt_embeds)
@@ -155,7 +163,7 @@ def infer(
             height=height,
             num_inference_steps=num_inference_steps,
             guidance_scale=guidance_scale,
-            controlnet_conditioning_scale=ip_adapter_strength, # ???????????????????????????????????????????????????????????????
+            controlnet_conditioning_scale=1.0 #ip_adapter_strength, 
             generator=generator
         ).images[0]
     else:
@@ -200,7 +208,6 @@ def infer(
             #strength_sn = float(strength_sn)
         
             # Используем ControlNet с LoRA
-            #pipe = pipe_controlnet
             prompt_embeds = long_prompt_encoder(prompt, pipe_controlnet.tokenizer, pipe_controlnet.text_encoder)
             negative_prompt_embeds = long_prompt_encoder(negative_prompt, pipe_controlnet.tokenizer, pipe_controlnet.text_encoder)
             prompt_embeds, negative_prompt_embeds = align_embeddings(prompt_embeds, negative_prompt_embeds)