Upload model_utils.py

src/model_utils.py

@@ -1,9 +1,22 @@
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+import torch
 
 def load_hf_model(model_name, device="cpu"):
     tokenizer = AutoTokenizer.from_pretrained(model_name)
     model = AutoModelForCausalLM.from_pretrained(model_name)
-    return pipeline("text-generation", model=model, tokenizer=tokenizer, device=0 if device=="cuda" else -1)
+    
+    # Handle meta tensors properly
+    if device == "cuda" and torch.cuda.is_available():
+        try:
+            model = model.to(device)
+        except NotImplementedError:
+            # If meta tensor error occurs, use to_empty()
+            model = model.to_empty(device=device)
+        device_id = 0
+    else:
+        device_id = -1
+    
+    return pipeline("text-generation", model=model, tokenizer=tokenizer, device=device_id)
 
 def generate_answer(text_gen, question, context):
     prompt = f"Context: {context}\n\nQuestion: {question}\n\nAnswer:"