Data Structures

Data Structures

Generated with Grok Build: Grok 4.5 · xAI Imagine (/imagine)

Module: src/types.jl

SpikeBatch

struct SpikeBatch
    spikes::Any                                    # trains / matrices / nested vectors
    times::Union{Nothing, Vector{Float32}}
    targets::Any                                   # optional supervised / teacher targets
end

Flexible typing (Any) so callers can use dense Float32 arrays, sparse formats, or custom containers. Timestamps optional when time is implicit in the array axis.

TraceBatch

struct TraceBatch
    traces::Any
end

Container for eligibility traces used by local learning rules.

TrainingState

mutable struct TrainingState
    loss::Float32
    metrics::Dict{Symbol, Any}
    traces::Union{Nothing, TraceBatch}
    gradients::Any
end

Constructed by train_step! after a step (at minimum loss + gradients from Zygote when AD succeeds).

Related

• Training Pipeline

---

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