Update app.py

app.py

@@ -8,6 +8,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import os
 import logging
 
 from threading import Thread
@@ -137,13 +138,15 @@ else:
     model = AutoModelForCausalLM.from_pretrained(
         "CohereForAI/c4ai-command-r-v01-4bit", 
         torch_dtype=torch.float16, 
-        device_map="auto"
+        device_map="auto",
+        token=os.getenv("HF_TOKEN")
     )
 
     device = model.device
 
     tokenizer = AutoTokenizer.from_pretrained(
-        "CohereForAI/c4ai-command-r-v01-4bit"
+        "CohereForAI/c4ai-command-r-v01-4bit",
+        token=os.getenv("HF_TOKEN")
     )
 
     tokenizer.use_default_system_prompt = False