Overview
Generated with Grok Build: Grok 4.5 · xAI Imagine (/imagine)
Wiki map for Limen-Neural/SynapticDistill.jl, grounded in package main @ e986a66 (v0.2.0).
Devin DeepWiki (structure reference, same commit):
https://app.devin.ai/org/your-organization-d33af45a/wiki/Limen-Neural/SynapticDistill.jl?branch=main
DeepWiki: 28 pages, generated 2026-07-26, locale en.
Online SNN training glue in Julia: inject model step + loss, run train_step!, use surrogates and (evolving) local learning rules — for pure SNNs or hybrid distillation.
At tip e986a66:
-
train_step!differentiates through the injected model step via Zygote and records loss/gradients inTrainingState. -
:eprop/:otttrule bodies are placeholders (log + dummy paths; not production-complete). -
predictive_eprop.jlis an empty placeholder (not exported). -
FPGA / Q8.8 export lives in
scripts/spikenaut_train.jl, not the package public API.
See Status and Roadmap.
| DeepWiki theme | GitHub wiki |
|---|---|
| Overview / getting started / architecture | Home, Overview, Getting Started, Architecture |
| Data structures | Data Structures |
| Training pipeline / train_step! | Training Pipeline |
| Learning rules / surrogates | Learning Rules, Surrogate Gradients |
| Hybrid MoE / examples / scripts | Hybrid Distillation, Examples and Scripts |
| Benchmarks / tests / license | Benchmarks, Testing, CI |
| Glossary | Glossary |
| Page | Purpose |
|---|---|
| Home | Intro |
| Getting Started | Install |
| Architecture | Layout |
| Data Structures | Types |
| Training Pipeline | train_step! |
| Learning Rules | Rules |
| Surrogate Gradients | Surrogates |
| Hybrid Distillation | Teacher–student |
| Examples and Scripts | Repo tools |
| Benchmarks | Perf harness |
| Testing / CI and Quality | Quality |
| Ecosystem | Siblings |
| Status and Roadmap | Honesty about stubs |
| Glossary | Terms |
Last updated: July 29, 2026
Updated by: Grok Build: Grok 4.5
Package tip reference: e986a66 (main, v0.2.0)
Devin DeepWiki: commit e986a66 · SynapticDistill.jl