removed `@gr.cache` decorator

app.py

@@ -14,26 +14,20 @@ CREATIVITY_SETTINGS = {
     2: {"top_p": 0.6, "temperature": 1.5},  # Creative
 }
 
-# Load model and tokenizer
-@gr.cache
-def load_model():
-    print("Loading model...")
-    base_model = BartForConditionalGeneration.from_pretrained(MODEL_NAME)
-    model = PeftModel.from_pretrained(base_model, PEFT_MODEL_ID)
-    tokenizer = BartTokenizer.from_pretrained(PEFT_MODEL_ID)
-    
-    # Set to eval mode
-    model.eval()
-    
-    # Move to GPU if available
-    device = "cuda" if torch.cuda.is_available() else "cpu"
-    model.to(device)
-    
-    print(f"Model loaded on {device}")
-    return model, tokenizer, device
+# Load model and tokenizer at startup (automatically cached)
+print("Loading model...")
+base_model = BartForConditionalGeneration.from_pretrained(MODEL_NAME)
+model = PeftModel.from_pretrained(base_model, PEFT_MODEL_ID)
+tokenizer = BartTokenizer.from_pretrained(PEFT_MODEL_ID)
+
+# Set to eval mode
+model.eval()
+
+# Move to GPU if available
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model.to(device)
 
-# Initialize model
-model, tokenizer, device = load_model()
+print(f"Model loaded on {device}")
 
 # Translation function
 def translate_to_emoji(text, creativity_level=1):