JoyNexus presents a unified multi-tenant service for post-training Vision-Language-Action models. It separates Training Model, Inference Model, and Environment Services behind APIs, using resident shared base models with tenant-specific slots. Tenants can run supervised fine-tuning, reinforcement learning, rollout, and evaluation through semantic or lower-level interfaces. Global training and inference queues schedule concurrent workloads while isolating action modules, optimizers, rollout records, and policy versions. The system also introduces group batching for heterogeneous VLA schemas with compatible model-facing prefixes, allowing one shared backbone forward pass across grouped samples. Evaluation uses workload simulation and an embodied-scenario batching pipeline.
No heat snapshots are available in the last 24 hours.