Add open_bench.py CLI for pipeline benchmarking

[Copilot]

The repository lacked a command-line tool for running benchmarks. Users had no way to benchmark pipelines without the web UI.

Changes

New CLI tool (open_bench.py)

• list: Enumerate available pipelines with status
• benchmark: Find max stream count at target FPS with configurable device/model/parameters
• optimize: Grid search over parameter combinations

Documentation (docs/user-guide/how-to-run-open-bench.md)

• Setup, usage examples, troubleshooting
• Docker and scripting workflows

Updated README.md

• Added CLI benchmarking to key features
• Linked to documentation

Example Usage

# List pipelines
python open_bench.py list

# Benchmark on CPU
python open_bench.py benchmark \
    --pipeline smartnvr \
    --video /path/to/video.mp4 \
    --device CPU \
    --fps-floor 30

# Optimize parameters on GPU
python open_bench.py optimize \
    --pipeline smartnvr \
    --video /path/to/video.mp4 \
    --device GPU \
    --batch-sizes 1 2 4 8 \
    --nireqs 1 2 4

The tool integrates with existing Benchmark and PipelineOptimizer classes, using pipeline configs from config.yaml for model defaults.

Original prompt

how to run this open_bench.py

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.