diff --git a/Dockerfile b/Dockerfile
index b77db07..5105496 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -46,5 +46,5 @@ RUN --mount=type=secret,id=HF_TOKEN,required=false \
     python3 /src/download_model.py; \
     fi
 
-
+# Start the handler
 CMD ["python3", "/src/handler.py"]
diff --git a/src/handler.py b/src/handler.py
index 4e6982c..176ec7e 100644
--- a/src/handler.py
+++ b/src/handler.py
@@ -3,20 +3,6 @@ import runpod
 from utils import JobInput
 from engine import vLLMEngine, OpenAIvLLMEngine
 
-# Detect number of visible GPUs
-gpu_count = torch.cuda.device_count()
-
-# Fallback to 1 if none detected
-if gpu_count < 1:
-    gpu_count = 1
-
-# Set the environment variable
-os.environ["TENSOR_PARALLEL_SIZE"] = str(gpu_count)
-
-print(f"Detected {gpu_count} GPU(s)")
-print(f"Set TENSOR_PARALLEL_SIZE={os.environ['TENSOR_PARALLEL_SIZE']}")
-
-
 vllm_engine = vLLMEngine()
 OpenAIvLLMEngine = OpenAIvLLMEngine(vllm_engine)