Update app.py

app.py

@@ -6,6 +6,15 @@ import os
 from transformers import AutoTokenizer, AutoModelForCausalLM
 from peft import PeftModel, PeftConfig
 
+# ✅ Retrieve Hugging Face token from environment variable (Secret)
+HF_TOKEN = os.getenv("HF_TOKEN")  # ✅ Now using stored secret, not hardcoded!
+
+if HF_TOKEN is None:
+    raise ValueError("❌ ERROR: Hugging Face token is missing! Please set HF_TOKEN in Hugging Face Secrets.")
+
+# ✅ Authenticate with Hugging Face
+login(token=HF_TOKEN)
+
 # ✅ Ensure offload directory exists
 os.makedirs("offload", exist_ok=True)
 
@@ -26,7 +35,8 @@ def load_model(model_path):
         base_model_name,
         torch_dtype=torch.float16,  # Use FP16 to reduce memory usage
         device_map="auto",  # Automatically distribute across CPU/GPU
-        offload_folder="offload"  # ✅ Prevents memory crashes
+        offload_folder="offload",  # ✅ Prevents memory crashes
+        use_auth_token=HF_TOKEN  # ✅ Authenticate model loading
     )
 
     # ✅ Load LoRA adapter with `is_trainable=False`