diff --git a/06_gpu_and_ml/vllm_generic.py b/06_gpu_and_ml/vllm_generic.py
index 96090d182..c4d1af34b 100644
--- a/06_gpu_and_ml/vllm_generic.py
+++ b/06_gpu_and_ml/vllm_generic.py
@@ -4,8 +4,7 @@
 
 import time
 
-from modal import Image, Stub, method, Secret
-
+from modal import Image, Stub, method
 
 vllm_image = (
     Image.from_registry(
@@ -25,9 +24,9 @@
 )
 class Model:
     def __init__(self, model_name: str):
+        import torch
         from vllm.engine.arg_utils import AsyncEngineArgs
         from vllm.engine.async_llm_engine import AsyncLLMEngine
-        import torch
 
         n_gpus = torch.cuda.device_count()
 
diff --git a/06_gpu_and_ml/vllm_generic_client.py b/06_gpu_and_ml/vllm_generic_client.py
index 6aec221d5..a89c9cd64 100644
--- a/06_gpu_and_ml/vllm_generic_client.py
+++ b/06_gpu_and_ml/vllm_generic_client.py
@@ -15,14 +15,12 @@
     secrets=[hf_secret],
     gpu=modal.gpu.A100(memory=40),
     volumes={"/hf-cache": model_vol},
-    allow_background_volume_commits=True,
 )
 
 Model_80GB = Model.with_options(
     secrets=[hf_secret],
     gpu=modal.gpu.A100(memory=80),
     volumes={"/hf-cache": model_vol},
-    allow_background_volume_commits=True,
 )
 
 mistral7b = Model_40GB(model_name="mistralai/Mistral-7B-Instruct-v0.2")