treeboost 0.1.0

High-performance Gradient Boosted Decision Tree engine for large-scale tabular data
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
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154

[package]
name = "treeboost"
version = "0.1.0"
edition = "2021"
description = "High-performance Gradient Boosted Decision Tree engine for large-scale tabular data"
license = "Apache-2.0"
repository = "https://github.com/ml-rust/treeboost"
homepage = "https://github.com/ml-rust/treeboost"
documentation = "https://docs.rs/treeboost"
keywords = [
  "machine-learning",
  "gbdt",
  "gradient-boosting",
  "decision-tree",
  "xgboost",
]
categories = ["science", "algorithms", "command-line-utilities"]
readme = "README.md"
authors = ["Farhan Syah"]
exclude = [
  "benchmarks/**",
  "benches/**",
  "catboost_info/**",
  "images/**",
  "results/**",
  "samples/**",
  "scripts/**",
  "target/**",
]

[lib]
name = "treeboost"
path = "src/lib.rs"
crate-type = ["cdylib", "rlib"]

[features]
default = []
python = ["dep:pyo3", "dep:numpy"]
gpu = ["dep:wgpu", "dep:pollster"]
cuda = ["dep:cudarc"]
mmap = ["dep:memmap2"]

[[bin]]
name = "treeboost"
path = "src/main.rs"

[dependencies]
# Data processing - columnar storage with lazy evaluation
polars = { version = "0.51", features = ["parquet", "lazy", "dtype-full"] }

# Parallelism - work-stealing scheduler
rayon = "1.10"

# Serialization - zero-copy model loading
rkyv = "0.8"

# Quantile estimation - streaming T-Digest for bin boundaries
tdigest = "0.2"

# Linear algebra - pure Rust, no C/Fortran linking
faer = "0.23"

# Safe transmutation for histogram entries
bytemuck = { version = "1.19", features = ["derive"] }

# Fast hashing for target encoding maps
rustc-hash = "2.1"

# Random number generation for data splitting
rand = "0.8"

# Error handling
thiserror = "2.0"

# Ordered floats for bin boundaries
ordered-float = "4.5"

# CLI argument parsing
clap = { version = "4.5", features = ["derive"] }

# Serialization framework
serde = { version = "1.0", features = ["derive"] }

# JSON serialization for predictions output
serde_json = "1.0"

# CSV serialization for tuner logging
csv = "1.3"

# Binary serialization for model files
bincode = "1.3"

# CRC32 for integrity checks in .trb files
crc32fast = "1.5"

# File locking for concurrent access prevention
fs4 = "0.13"

# Timestamp for run directories
chrono = "0.4"

# Python bindings (optional)
pyo3 = { version = "0.23", features = ["extension-module"], optional = true }

# NumPy support for Python bindings
numpy = { version = "0.23", optional = true }

# GPU compute via WebGPU (optional - for gpu feature)
wgpu = { version = "27", optional = true }

# Async executor for blocking WGPU calls (optional - for gpu feature)
pollster = { version = "0.4", optional = true }

# CUDA driver bindings (optional - for cuda feature)
cudarc = { version = "=0.18.2", optional = true, default-features = false, features = [
  "driver",
  "nvrtc",
  "dynamic-linking",
  "cuda-version-from-build-system",
] }

# Memory-mapped file I/O (optional - for mmap feature)
memmap2 = { version = "0.9", optional = true }

[dev-dependencies]
criterion = { version = "0.5", features = ["html_reports"] }
approx = "0.5"
tempfile = "3.15"
fastrand = "2.3"

# Competitor GBDT implementations for benchmarking
gbdt = { version = "0.1", features = ["enable_training"] }
forust-ml = "0.4"

[[bench]]
name = "competitors"
harness = false

[[bench]]
name = "profile"
harness = false

[[bench]]
name = "correctness"
harness = false

[profile.release]
lto = "thin"
codegen-units = 1
opt-level = 3

[profile.bench]
lto = "thin"
codegen-units = 1