Add GuideLLM-based vLLM Benchmarking Script

Overview

Added run_guidellm.sh script for benchmarking vLLM inference servers using the official GuideLLM tool.

💾 Result Management

• Timestamped output directories: results_<IP>_<TIMESTAMP>/
• Multiple formats: JSON (detailed), CSV (tabular), HTML (visual)
• Prevents result overwriting across multiple runs
• Results stored in ~/guidellm_bench/

Usage Examples

Basic Synthetic Benchmark:

./run_guidellm.sh --ip 99.79.191.224

HuggingFace Dataset:

./run_guidellm.sh --ip 99.79.191.224 \
  --data "abisee/cnn_dailymail" \
  --data-args '{"name":"3.0.0"}' \
  --data-column-mapper '{"text_column":"article"}' \
  --data-samples 1000