Update app.py

app.py

@@ -12,6 +12,8 @@ import requests
 from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
 from qwen_vl_utils import process_vision_info
 
+from diffusers import DiffusionPipeline
+
 
 
 
@@ -33,19 +35,23 @@ processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-larg
 model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
 processor1 = BlipProcessor.from_pretrained("noamrot/FuseCap")
 model2 = BlipForConditionalGeneration.from_pretrained("noamrot/FuseCap")
-pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-3.5-medium")
+# pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-3.5-medium")
 # model3 =model = Qwen2VLForConditionalGeneration.from_pretrained(
 #     "prithivMLmods/Qwen2-VL-OCR-2B-Instruct", torch_dtype="auto", device_map="auto"
 # )
 # processor2 = AutoProcessor.from_pretrained("prithivMLmods/Qwen2-VL-OCR-2B-Instruct")
 
+pipe3 = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-dev")
+pipe3.load_lora_weights("tryonlabs/FLUX.1-dev-LoRA-Outfit-Generator")
+
 
 
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
-pipe.to(device)
+# pipe.to(device)
 model2.to(device)
 model.to(device)
+pip3.to(device)
 
 
 
@@ -107,10 +113,10 @@ def generate_caption_and_image(image):
 
     prompt = f'''Create a highly realistic clothing item based on the following descriptions: The design should reflect {caption1} and {caption2}, blending both themes into a single, stylish, and modern piece of clothing. Incorporate highly realistic and high-quality textures that exude sophistication, with realistic fabric lighting and fine details. Subtly hint at {selected_fabric}, featuring a {selected_pattern} motif and a {selected_textile_design} style that harmoniously balances the essence of both captions.'''
 
-
+  ]
 
     # Generate image based on the caption
-    generated_image = pipe(prompt).images[0]
+    generated_image = pipe3(prompt).images[0]
 
     return prompt, generated_image