Trajectory Labs releases C-LoRA, an open-source training stack with 2.81x higher throughput and no reward degradation

The framework includes SkyRL, developed in partnership with NovaSkyAI.