diff --git a/scripts/olmo2_modal_openai.py b/scripts/olmo2_modal_openai.py
index 570d928ed..ab21e31ad 100644
--- a/scripts/olmo2_modal_openai.py
+++ b/scripts/olmo2_modal_openai.py
@@ -33,6 +33,7 @@
 # the weights from HuggingFace directly into a local directory when building the
 # container image.
 
+
 def download_model_to_image(model_dir, model_name, model_revision):
     from huggingface_hub import snapshot_download
     from transformers.utils import move_cache
@@ -46,6 +47,7 @@ def download_model_to_image(model_dir, model_name, model_revision):
     )
     move_cache()
 
+
 # ## Set up the container image
 
 # Our first order of business is to define the environment our server will run in
@@ -76,7 +78,7 @@ def download_model_to_image(model_dir, model_name, model_revision):
     .env({"HF_HUB_ENABLE_HF_TRANSFER": "1"})
     .run_function(
         download_model_to_image,
-        timeout=60 * MINUTES, # typically much faster but set high to be conservative
+        timeout=60 * MINUTES,  # typically much faster but set high to be conservative
         kwargs={
             "model_dir": MODEL_DIR,
             "model_name": MODEL_NAME,
@@ -100,14 +102,15 @@ def download_model_to_image(model_dir, model_name, model_revision):
 #
 app = modal.App(APP_NAME)
 
+
 @app.function(
     image=vllm_image,
     gpu=GPU_CONFIG,
-    keep_warm=0, # Spin down entirely when idle
+    keep_warm=0,  # Spin down entirely when idle
     container_idle_timeout=5 * MINUTES,
     timeout=24 * HOURS,
     allow_concurrent_inputs=1000,
-    secrets=[modal.Secret.from_name("example-secret-token")], # contains MODAL_TOKEN used below
+    secrets=[modal.Secret.from_name("example-secret-token")],  # contains MODAL_TOKEN used below
 )
 @modal.asgi_app()
 def serve():
@@ -144,7 +147,7 @@ def serve():
 
     # This example uses a token defined in the Modal secret linked above,
     # as described here: https://modal.com/docs/guide/secrets
-    async def is_authenticated(api_key = fastapi.Security(http_bearer)):
+    async def is_authenticated(api_key=fastapi.Security(http_bearer)):
         if api_key.credentials != os.getenv("MODAL_TOKEN"):
             raise fastapi.HTTPException(
                 status_code=fastapi.status.HTTP_401_UNAUTHORIZED,
@@ -167,17 +170,13 @@ async def is_authenticated(api_key = fastapi.Security(http_bearer)):
         enforce_eager=False,  # capture the graph for faster inference, but slower cold starts (30s > 20s)
     )
 
-    engine = AsyncLLMEngine.from_engine_args(
-        engine_args, usage_context=UsageContext.OPENAI_API_SERVER
-    )
+    engine = AsyncLLMEngine.from_engine_args(engine_args, usage_context=UsageContext.OPENAI_API_SERVER)
 
     model_config = get_model_config(engine)
 
     request_logger = RequestLogger(max_log_len=2048)
 
-    base_model_paths = [
-        BaseModelPath(name=MODEL_NAME.split("/")[1], model_path=MODEL_NAME)
-    ]
+    base_model_paths = [BaseModelPath(name=MODEL_NAME.split("/")[1], model_path=MODEL_NAME)]
 
     api_server.chat = lambda s: OpenAIServingChat(
         engine,