Update app.py

app.py

@@ -8,7 +8,6 @@ from diffusers import DiffusionPipeline
 import torch
 import spaces  # Hugging Face Spaces module
 
-from transformers import VisionEncoderDecoderModel, ViTImageProcessor, AutoTokenizer
 
 
 
@@ -24,15 +23,12 @@ login(token=hf_token)
 processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
 model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
 
-model2 = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
-feature_extractor = ViTImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
-tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
 
 
 pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-3.5-medium")
 device = "cuda" if torch.cuda.is_available() else "cpu"
 pipe.to(device)
-model2.to(device)
+
 model.to(device)
 
 
@@ -54,21 +50,15 @@ def generate_caption_and_image(image):
     selected_fabric = random.choice(fabrics)
     selected_pattern = random.choice(patterns)
     selected_textile_design = random.choice(textile_designs)
-    
-    gen_kwargs = {"max_length": 16, "num_beams": 4}
-
-
-    pixel_values = feature_extractor(images=[img], return_tensors="pt").pixel_values.to(device)
-    output_ids = model.generate(pixel_values, **gen_kwargs)
-    caption1 = tokenizer.decode(output_ids[0], skip_special_tokens=True).strip()
 
 
+    caption2 =""
     
     # Generate caption
     inputs = processor(image, return_tensors="pt", padding=True, truncation=True, max_length=250)
     inputs = {key: val.to(device) for key, val in inputs.items()}
     out = model.generate(**inputs)
-    caption2 = processor.decode(out[0], skip_special_tokens=True)
+    caption1 = processor.decode(out[0], skip_special_tokens=True)
 
     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.'''