site_name: miniVERL

site_description: Auditable one-GPU online post-training and a pinned verl scale-out bridge

site_url: https://daoyuanli2816.github.io/mini-verl/

repo_url: https://github.com/DaoyuanLi2816/mini-verl

repo_name: DaoyuanLi2816/mini-verl

theme:

name: readthedocs

nav:

- Home: index.md

- One-GPU quickstart: single-gpu-guide.md

- Alignment Lab:

- Why OPD after SFT?: alignment-lab/when-opd-should-follow-sft.md

- SFT vs DPO vs OPD: alignment-lab/alignment-lab-v1.md

- Consumer runtime: consumer-runtime-v1.md

- RecoveryBench: recoverybench/recoverybench-v1.md

- Verified verl bridge: verl-bridge.md

- 90-second bridge demo: verl-bridge-demo.md

- Community benchmarks: community-benchmarks.md

- v0.6 launch story: verl-bridge-launch.md

- Schemas:

- Trajectories: trajectory-schema.md

- Benchmarking: benchmarking.md

- Limitations: limitations.md

markdown_extensions:

- tables

- fenced_code

- admonition

- toc:

permalink: true