push model cash path

app.py

@@ -29,7 +29,7 @@ model_cache_path = snapshot_download(
     local_dir="/app/model_cache"            # <-- direct destination
 )
 
-print(">>> MODEL CACHE PATH:", MODEL_PATH, os.listdir(MODEL_PATH))
+print(">>> MODEL CACHE PATH:", model_cache_path, os.listdir(model_cache_path))
 
 device = "gpu" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
 dtype = torch.float16 if device in ("mps", "gpu") else torch.float32
@@ -50,18 +50,18 @@ def load_model():
     # model_name, torch_dtype=dtype, trust_remote_code=True
     # )
     model = AutoModelForCausalLM.from_pretrained(
-        MODEL_PATH,
+        model_cache_path,
         local_files_only=True,
         torch_dtype=dtype,
         trust_remote_code=True
     ).to(device).eval()
     # tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
     tokenizer = AutoTokenizer.from_pretrained(
-        MODEL_PATH,
+        model_cache_path,
         local_files_only=True,
         trust_remote_code=True
     )
-    print("✅ Model and tokenizer loaded from", MODEL_PATH)
+    print("✅ Model and tokenizer loaded from", model_cache_path)
 
 def predict_NuExtract(texts, template, batch_size=10, max_length=5096, max_new_tokens=1024):
     print("Starting NuExtract prediction...", flush=True)