Update trtllm_llama.py (#797)

I was testing this file in modal examples and it broke. This fixes the error `TypeError: 'GenerateRequest' object is not subscriptable`.

06_gpu_and_ml/llm-serving/trtllm_llama.py

@@ -576,7 +576,7 @@ class GenerateRequest(pydantic.BaseModel):
 )
 def generate_web(data: GenerateRequest) -> list[str]:
     """Generate responses to a batch of prompts, optionally with custom inference settings."""
-    return Model.generate.remote(data["prompts"], settings=None)
+    return Model.generate.remote(data.prompts, settings=None)
 
 
 # To set our function up as a web endpoint, we need to run this file --