god-graph 0.6.0-alpha

A graph-based LLM white-box optimization toolbox: topology validation, Lie group orthogonalization, tensor ring compression
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

//! Tensor 模块：为图神经网络和高性能计算提供张量支持
//!
//! 本模块实现了 God-Graph 的 tensor 基础设施，包括：
//! - Dense tensor（密集张量）：基于 ndarray 的 N 维数组
//! - Sparse tensor（稀疏张量）：COO, CSR, BSR 格式
//! - Tensor 操作：矩阵乘法、转置、归约等
//! - 多后端支持：NdArray, Dfdx (GPU), Candle
//! - 内存池优化：减少分配开销
//! - 梯度检查点：降低反向传播内存占用
//!
//! ## 特性
//!
//! - **后端抽象**：通过 trait 系统支持多种 backend（ndarray, dfdx, candle）
//! - **稀疏格式**：COO（坐标格式）、CSR（压缩稀疏行）、BSR（块稀疏行）
//! - **SIMD 优化**：使用 wide crate 实现 SIMD 向量化
//! - **内存对齐**：64 字节缓存行对齐，避免 false sharing
//! - **内存池**：可复用的张量分配，适用于迭代算法
//!
//! ## 示例
//!
//! ```
//! # #[cfg(feature = "tensor")]
//! # {
//! use god_gragh::tensor::{DenseTensor, TensorBase};
//!
//! // 创建 2x3 密集张量
//! let data = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0];
//! let tensor = DenseTensor::from_vec(data, vec![2, 3]);
//!
//! assert_eq!(tensor.shape(), &[2, 3]);
//! assert_eq!(tensor.ndim(), 2);
//! # }
//! ```

#[cfg(feature = "tensor")]
pub mod traits;

#[cfg(feature = "tensor")]
pub mod dense;

#[cfg(feature = "tensor")]
pub mod sparse;

#[cfg(feature = "tensor")]
pub mod ops;

#[cfg(feature = "tensor")]
pub mod error;

#[cfg(feature = "tensor")]
pub mod types;

#[cfg(feature = "tensor")]
pub mod backend;

#[cfg(feature = "tensor-pool")]
pub mod pool;

#[cfg(feature = "tensor-gnn")]
pub mod gnn;

#[cfg(feature = "tensor")]
pub mod graph_tensor;

#[cfg(feature = "tensor")]
pub mod differentiable;

#[cfg(feature = "tensor")]
pub mod decomposition;

#[cfg(feature = "tensor")]
pub mod unified_graph;

// 重新导出核心类型
#[cfg(feature = "tensor")]
pub use traits::{COOView, DType, Device, SparseTensorOps, TensorBase, TensorOps};

#[cfg(feature = "tensor")]
pub use dense::DenseTensor;

#[cfg(feature = "tensor")]
pub use sparse::{COOTensor, CSRTensor, SparseTensor};

#[cfg(feature = "tensor")]
pub use error::TensorError;

#[cfg(feature = "tensor")]
pub use types::{EdgeFeatures, NodeFeatures, TensorEdge, TensorNode};

#[cfg(feature = "tensor")]
pub use types::AdjacencyMatrix;

#[cfg(feature = "tensor")]
pub use types::DegreeMatrix;

#[cfg(feature = "tensor")]
pub use backend::{NdArrayStorage, TensorStorage, UnifiedStorage};

#[cfg(feature = "tensor-pool")]
pub use pool::{ArenaStats, ArenaTensor, PoolConfig, PoolStats, PooledTensor, TensorArena, TensorPool};

#[cfg(feature = "tensor-autograd")]
pub use pool::GradientCheckpoint;

#[cfg(feature = "tensor-gnn")]
pub use gnn::{
    Aggregator, GATConv, GCNConv, GraphSAGE, IdentityMessage, LinearMessage, MaxAggregator,
    MeanAggregator, MessageFunction, MessagePassingLayer, SumAggregator, UpdateFunction,
};

#[cfg(feature = "tensor")]
pub use graph_tensor::{
    GraphFeatureExtractor, GraphReconstructor,
};

#[cfg(feature = "tensor")]
pub use graph_tensor::{
    GraphAdjacencyMatrix, GraphBatch, GraphTensorExt,
};

#[cfg(feature = "tensor")]
pub use differentiable::{
    DifferentiableEdge, DifferentiableGraph, DifferentiableNode, EdgeEditOp, EdgeEditPolicy,
    EditOperation, GradientConfig, GradientRecorder, GraphTransformer, GumbelSoftmaxSampler,
    NodeEditOp, StructureEdit, ThresholdEditPolicy,
};

#[cfg(feature = "tensor")]
pub use unified_graph::{EdgeData, NodeData, UnifiedConfig, UnifiedGraph};

#[cfg(feature = "tensor")]
pub use decomposition::{
    lie_algebra::{lie_exponential, lie_logarithm, skew_symmetric_projection},
    qr::{orthogonalize, qr_decompose},
    svd::{low_rank_approx, svd_decompose},
    tensor_ring::{compress_tensor_ring, tensor_ring_decompose, TensorRing},
};