Update app.py

app.py

@@ -4,57 +4,34 @@ from PIL import Image
 import whisper
 from transformers import AutoProcessor, AutoModelForImageTextToText
 
-processor = AutoProcessor.from_pretrained("deepseek-community/Janus-Pro-1B",trust_remote_code=True)
-model = AutoModelForImageTextToText.from_pretrained("deepseek-community/Janus-Pro-1B",trust_remote_code=True)
+
+processor = AutoProcessor.from_pretrained("deepseek-community/Janus-Pro-1B", trust_remote_code=True)
+model = AutoModelForImageTextToText.from_pretrained("deepseek-community/Janus-Pro-1B", trust_remote_code=True)
 whisper_model = whisper.load_model("base")
 
 def build_instruction(user_text):
-    return f"""
-You are a professional AI prompt engineer.
-
-Convert the input into a highly detailed AI generation prompt.
-
-Include:
-- Subject
-- Environment
-- Summary
-
-Make it visually rich and optimized for all existing AI models.
-The Prompt should e detailed prompt about words ranging from 50 to 70.
+    return f"You are a professional AI prompt engineer. Convert the input into a highly detailed AI generation prompt. Include: Subject, Environment, Summary. Input: {user_text}\nReturn only the final prompt."
 
-Input: {user_text}
-
-Return only the final prompt.
-"""
 def text_to_prompt(user_text):
     instruction = build_instruction(user_text)
-    
-    inputs = processor(
-        text=instruction,
-        return_tensors="pt"
-    ).to(model.device)
+    inputs = processor(text=instruction, return_tensors="pt")
 
     input_len = inputs.input_ids.shape[1]
-    output = model.generate(**inputs, max_new_tokens=150) 
-    generated_tokens = output[0][input_len:]
-
-    return processor.decode(generated_tokens, skip_special_tokens=True)
+    
+    output = model.generate(**inputs, max_new_tokens=150)
+    
+    return processor.decode(output[0][input_len:], skip_special_tokens=True)
 
 def image_text_to_prompt(image_path, user_text):
     image = Image.open(image_path)
     instruction = build_instruction(user_text)
-
-    inputs = processor(
-        images=image,
-        text=instruction,
-        return_tensors="pt"
-    ).to(model.device)
-
+    inputs = processor(images=image, text=instruction, return_tensors="pt")
+    
     input_len = inputs.input_ids.shape[1]
+    
     output = model.generate(**inputs, max_new_tokens=150)
-    generated_tokens = output[0][input_len:]
-
-    return processor.decode(generated_tokens, skip_special_tokens=True)
+  
+    return processor.decode(output[0][input_len:], skip_special_tokens=True)
 
 def audio_to_prompt(audio_path):
     result = whisper_model.transcribe(audio_path)