Update app.py

app.py

@@ -1,14 +1,21 @@
+import os
 import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 from threading import Thread
 
+# This pulls the token you just saved in the Space settings
+hf_token = os.getenv("HF_TOKEN")
+
 model_id = "ZyperAI/Z-AI-0.1-1.1B-Code.web"
-tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+# We pass the token to both the tokenizer and the model
+tokenizer = AutoTokenizer.from_pretrained(model_id, token=hf_token)
 model = AutoModelForCausalLM.from_pretrained(
     model_id, 
     torch_dtype=torch.float32, 
-    device_map="cpu"
+    device_map="cpu",
+    token=hf_token
 )
 
 def generate_code(prompt, history):