Linex is the all-in-one infra for Agentic RL - training, inference, rollout, sandbox layers combined without glue.
Linex features minimalism - optimized interfaces, boilerplate-free, light dependency, and the one-piece philosophy (deprecated recipe and model supports are dynamically dropped).
It's a power fork from Molt (Pytorch FSDP + glue-free Trainer) and Polar (efficient containerized agent rollout).
Linex requires Python 3.11 or newer:
python3 -m pip install -e .Skill2Env RL DPPO: asynchronous DPPO training with skill2env dataset.
