Update app_v4.py

app_v4.py

@@ -33,15 +33,20 @@ if device == "cuda:0":
 tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=False)
 
 # Attempt to load the model, catch any OOM errors
-model_loaded = False
-try:
-    model = AutoGPTQForCausalLM.from_quantized(
+@st.cache_data
+def load_gptq_model():
+    AutoGPTQForCausalLM.from_quantized(
         pretrained_model_dir,
         model_basename="Jackson2-4bit-128g-GPTQ",
         use_safetensors=True,
         device=device
     )
     model.eval()  # Set the model to inference mode
+    return model
+    
+model_loaded = False
+try:
+    model = load_gptq_model()
     model_loaded = True
 except RuntimeError as e:
     if 'CUDA out of memory' in str(e):