# Initialize the VLLM wrapper (model loads on first use)
The first call to `run()` will download and load the model.

# Initialize the VLLM wrapper (model loads on first use)
# The model is loaded when `run()` is first invoked

    # Model initialization is deferred until `run()` is called

# Initialize the vLLM wrapper (weights load on first run)
The model is loaded lazily when `run()` is invoked.

# The weights are loaded lazily when you first call `run()`

# Model weights are loaded during the first call to `run()`