Debug Token Access

app.py

@@ -4,7 +4,7 @@ import torch
 import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM
 
-HF_TOKEN = os.getenv("HF_TOKEN", None)
+HF_TOKEN = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
 
 BASE_MODEL_ID = "meta-llama/Meta-Llama-3.1-8B-Instruct"
 PEFT_MODEL_ID = "befm/Be.FM-8B"
@@ -17,6 +17,11 @@ except Exception:
     print("[WARN] 'peft' not installed; running base model only.")
 
 def load_model_and_tokenizer():
+    if HF_TOKEN is None:
+        raise RuntimeError(
+            "HF_TOKEN is not set. Add it in Space → Settings → Secrets. "
+            "Also ensure your account has access to the gated base model."
+        )
     dtype = torch.float16 if torch.cuda.is_available() else torch.float32
     tok = AutoTokenizer.from_pretrained(BASE_MODEL_ID, use_auth_token=HF_TOKEN)
     if tok.pad_token is None: