Start with the runnable quickstart. The cnskit API is the supported task-training interface; cns_tinker retains the earlier simulation and ridge-readout interfaces.
| Guide | Contents |
|---|---|
| Training | Episodes, splits, masks, adaptation modes and CLI |
| Learning API | Graph, ConnectomeModel, Trainer and checkpoint contracts |
| MaleCNS data | Official download, integrity checks, graph preparation and explicit currents |
| Validation | Reproducible evidence and its scope |
| Architecture | Sparse recurrence, trainable components and export format |
| Legacy API | Earlier Session and scenario interfaces |
| Ridge readouts | Lightweight NumPy baseline |
| Troubleshooting | Installation, memory and invalid inputs |
| Roadmap | Current limits and planned work |
| Licensing | Source and upstream data terms |