Update app.py

app.py

@@ -6,7 +6,8 @@ from huggingface_hub import InferenceClient
 For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
 """
 
-client = InferenceClient(model="HuggingFaceH4/zephyr-7b-beta", token="HUGGINGFACEHUB_API_TOKEN")
+import os
+client = InferenceClient(model="HuggingFaceH4/zephyr-7b-beta", token=os.getenv("HUGGINGFACEHUB_API_TOKEN"))
 
 
 def respond(
@@ -85,7 +86,9 @@ base_model = AutoModelForCausalLM.from_pretrained(
     base_model_name, 
     torch_dtype=torch.float16, 
     device_map="auto",
-    token=api_token  # ✅ Pass token here
+    use_auth_token=api_token  # ✅ Correct
+)
+
 )
 
 # Load fine-tuned model