feat: TensorRT-LLM Support for logits_prob

[hiro-v]

Hi there,

Currently I'm using cortex.tensorrt-llm in order to benchmark for MMLU and TruthfulQA using https://github.com/EleutherAI/lm-evaluation-harness/tree/main

However, the current API does not return logits_prob

Please add logits_prob in chat_completion API as it's supported from 0.8.0 https://github.com/NVIDIA/TensorRT-LLM/issues/983

Thank you

[gabrielle-ong]

Closing all open Tensorrt-llm stories due to TensorRT-LLM not supporting Desktop
Parent issue: menloresearch/cortex.cpp#1742