katgpt-types 0.1.1

Shared configuration, RNG, math utilities, LoRA, domain embeddings, and inference types for katgpt-rs / riir-engine. Pure substrate leaf — no katgpt-* deps.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138

//! katgpt-types — Shared configuration, RNG, math utilities, SIMD kernels,
//! and inference types for the katgpt-rs / riir-engine superset.
//!
//! Pure substrate leaf crate. No katgpt-* dependencies — only `fastrand`,
//! `blake3`, `serde`, `half`. This is the foundational layer (types + SIMD
//! kernels) that every other katgpt-* crate (and riir-engine) builds on.
//! The `types` and `simd` modules are co-located because `types::math`
//! calls `simd` kernels (softmax / rmsnorm / matmul) and `simd::ternary`
//! uses `types::TernaryWeights` — they form a tight bidirectional leaf that
//! cannot be split further without breaking the cycle.
//!
//! Originally a single 5,148-line `types.rs` inside katgpt-core (2.5× the
//! 2048-line ceiling), split into topic-specific submodules. The full public
//! surface is re-exported here so consumers can use `katgpt_types::*` paths
//! directly.
//!
//! Spun out of `katgpt-core::types` (Issue 007 Phase E Tier 1 #2) as a
//! standalone publishable crate mirroring the `katgpt-dec` / `katgpt-transformer`
//! template.
//!
//! # Module layout
//!
//! - [`enums`] — small config enums (DepthTier, HlaMode, AttentionMode, …)
//!   plus WallConfig / ThinkingBudget
//! - [`config`] — the `Config` struct (~1.5k lines), `InferenceOverrides`,
//!   and `kv_dim`
//! - [`rng`] — XorShift64 PRNG
//! - [`math`] — SIMD-accelerated softmax / rmsnorm / matmul / sample_token
//!   (legacy home; candidates for relocation to `katgpt-simd::`)
//! - [`lora`] — CPU-side LoRA adapter
//! - [`gpart`] — GPart Isometric Partition adapter (Research 227)
//! - [`domain`] — DomainLatent embedding (Plan 038)
//! - [`inference`] — InferenceResult, TaskType, ProposerTask, DataGate
//! - [`looping`] — Training-Free Loop types (Plan 136)
//! - [`ternary`] — Bit-plane ternary weights (`plasma_path`)
//! - [`hydra`] — Hydra Adaptive Layer Budget types
//! - [`sense`] — ShardEmbedding + sense composition types
//!
//! Test modules live alongside their topic (e.g. `rng::tests_rng`) or in
//! `tests_types.rs` for cross-cutting tests.

mod config;
/// Depth-Invariance Diagnostic & Magnitude-Regularized Residual (Plan 306).
/// Pure math, depends only on simd. Co-located here (the leaf) so both
/// katgpt-core (depth_invariance feature) and katgpt-micro-belief
/// (audit_depth_invariance methods) can consume it without a cycle.
#[cfg(feature = "depth_invariance")]
pub mod depth_invariance;
mod domain;
#[cfg(feature = "depth_invariance")]
pub use depth_invariance::{
    DepthInvarianceConfig, DepthInvarianceDiagnostic, DepthInvarianceKind, MagnitudeRegularization,
    Scratch, apply_magnitude_regularization, classify_chain, classify_chain_batched,
};
mod enums;
mod gpart;
mod hydra;
mod inference;
/// QuantizedKVCache trait — shared extension point for all KV backends.
/// Promoted from `katgpt-rs/src/types.rs` (Issue 015 Phase 1) so backend
/// crates can depend on the trait without pulling the whole root crate.
pub mod kv_cache;
/// Shared leaky-integrator / delta-rule step primitive (Plan 276 Phase 2 T2.1).
/// Pure inline math, zero deps. Consumed by both katgpt-micro-belief
/// (`LeakyIntegrator::step`) and katgpt-core's sense reconstruction
/// (`ReconstructionState::evolve_hla`). Co-located here (the leaf) so both
/// consumers can share the single source of truth without a cycle.
pub mod leaky_core;
mod looping;
mod lora;
pub mod math;
pub mod merkle;
mod rng;
mod sense;
/// SIMD-accelerated linear algebra kernels (NEON / AVX2 / WASM-SIMD128 /
/// scalar fallback). Co-located with `types` because `types::math` calls
/// these kernels and `simd::ternary` uses `types::TernaryWeights`.
pub mod simd;
pub mod slod;
pub mod temporal;
mod ternary;

#[cfg(test)]
mod tests_types;

// Re-export the entire public surface so `katgpt_types::*` paths are
// available at the crate root. Feature gates mirror the gates on the
// underlying items.
pub use config::{Config, InferenceOverrides, kv_dim};
#[cfg(feature = "domain_latent")]
pub use domain::DomainLatent;
#[cfg(feature = "deltanet_inference")]
pub use enums::DeltaNetLayerType;
#[cfg(feature = "collapse_aware_thinking")]
pub use enums::ThinkingBudget;
#[cfg(feature = "wall_attention")]
pub use enums::WallConfig;
pub use enums::{
    AttentionMode, AttentionProjection, CacheLayout, CalibrationMode, ConvergenceSelector,
    DashAttnConfig, DepthTier, HlaMode, HybridPattern, LoopMode, ModelArchitecture, ResidualGate,
    RetrievalHeadRole, RtTurboConfig, SdpaOutputGate, WeightDtype,
};
#[cfg(feature = "sr2am_configurator")]
pub use enums::{ConfiguratorContext, PlanningDecision};
#[cfg(feature = "gpart_adapter")]
pub use gpart::{GPART_MAGIC, GPART_VERSION, GpartAdapter, GpartPair, GpartPrepared};
#[cfg(feature = "hydra_budget")]
pub use hydra::{HydraBudgetConfig, HydraLayerProfile};
pub use inference::InferenceResult;
#[cfg(feature = "data_gate")]
pub use inference::{DataGate, GateDecision, ProposerTask, TaskType};
pub use kv_cache::QuantizedKVCache;
pub use leaky_core::leaky_step;
pub use looping::{CacheStrategy, IterationMode, SubStepStrategy, TrainingFreeLoopConfig};
pub use lora::{LoraAdapter, LoraPair, lora_apply};
#[allow(deprecated)]
pub use math::sample_token;
#[cfg(feature = "sparse_mlp")]
pub use math::sparse_matmul;
pub use math::{
    gegelu, gegelu_tanh, matmul, matmul_f16, matmul_f16_parallel, matmul_parallel, matmul_relu,
    rmsnorm, rmsnorm_with_gamma, rmsnorm_with_gamma_eps, sample_token_into, silu, softmax,
    softmax_scaled, swiglu,
};
pub use merkle::{
    HASH_SIZE, MERKLE_OCTREE_BRANCHING, MERKLE_OCTREE_DEPTH, MERKLE_OCTREE_INTERNAL,
    MERKLE_OCTREE_LEAVES, MERKLE_OCTREE_NODES, MerkleOctree, MerkleProof,
};
pub use rng::Rng;
pub use sense::{DilationConfig, SenseKind, SenseModule, ShardEmbedding, TernaryDir};
pub use slod::ScaleBoundary;
pub use temporal::{TemporalDerivativeKernel, sigmoid_surprise_gate};
#[cfg(feature = "plasma_path")]
pub use ternary::TernaryWeights;

// Internal helpers (read_u32_le / read_f32_le / read_u16_le) live in
// `domain.rs` and are crate-private — not re-exported here. If other modules
// need them, import via `crate::domain::read_u32_le`.