bench plugin¶
Manage benchmark plugins. The core CLI ships with no plugins — install at least one (inferencebench-llm) before running benchmarks.
The shorthand bench plugins (note plural) is equivalent to bench plugin list.
Synopsis¶
bench plugin list¶
Expected output:
Installed plugins
Name Module Distribution
llm.inference inferencebench_llm.plugin:LlmInferencePlugin inferencebench-llm
With no plugins installed:
bench plugin init¶
Scaffold a new plugin package under ./plugins/<name>/ with a registered entry point, src package, a passing smoke test, and a README stub.
| Option | Default | Description |
|---|---|---|
--kind |
both |
perf, quality, or both. |
--modality |
"" |
One of llm, voice, video, 3d, etc. |
The plugin name must match [a-z][a-z0-9-]* (lowercase, no underscores).
bench plugin install¶
Phase 1 stub — until ticket 0028 lands, this command prints a hint and exits. Use pip install inferencebench-<name> directly.
bench plugin info¶
Expected output:
llm.inference
module: inferencebench_llm.plugin:LlmInferencePlugin
distribution: inferencebench-llm 0.0.0
summary: InferenceBench plugin for LLM inference performance.
homepage: https://github.com/yobitelcomm/bench
See also¶
- bench list — see the benchmarks each plugin exposes
- Plugins overview