evoc 0.0.1

Embedding Vector Oriented Clustering — fast clustering of high-dimensional embedding vectors (Rust port of EVōC)
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

[package]
name = "evoc"
version = "0.0.1"
edition = "2021"
rust-version = "1.70"
description = "Embedding Vector Oriented Clustering — fast clustering of high-dimensional embedding vectors (Rust port of EVōC)"
license = "BSD-2-Clause"
readme = "README.md"
documentation = "https://docs.rs/evoc"
homepage = "https://github.com/eugenehp/evoc-rs"
repository = "https://github.com/eugenehp/evoc-rs"
keywords = ["clustering", "embeddings", "knn", "hdbscan", "umap"]
categories = ["science", "algorithms"]
authors = ["Eugene Hauptmann"]
# Dev-only paths (parity fixtures, benchmarks, scripts) — not published to crates.io.
exclude = [
    "/.venv-parity",
    "/examples/output",
    "/examples/*.py",
    "/tests/fixtures",
    "/benches/history",
    "/scripts",
]

[package.metadata.docs.rs]
all-features = true

[dependencies]
ndarray = { version = "0.16", features = ["rayon"] }
rayon = "1.10"
rand = "0.8"
sprs = "0.11"
rustc-hash = "2"
thiserror = "2"
serde_json = { version = "1", optional = true }
faer = { version = "0.20", default-features = false, features = ["std", "svd", "evd"] }
ndarray-npy = { version = "0.9", optional = true }
rlx-cpu = { version = "0.2.1", optional = true }
rlx = { version = "0.2.1", optional = true, default-features = false }
ureq = { version = "2.12", default-features = false, features = ["gzip", "tls"], optional = true }
flate2 = { version = "1.1", optional = true }
tar = { version = "0.4", optional = true }
zip = { version = "2.2", default-features = false, features = ["deflate"], optional = true }

[features]
default = ["full"]
# Convenience: enable the full pipeline, `.npy` I/O, and bundled dataset loaders.
full = ["cluster", "npy", "datasets"]
# Pipeline stages (each implies the previous).
cluster = ["embed"]
embed = ["init"]
init = ["graph"]
graph = ["knn"]
knn = ["dep:rlx-cpu"]
# `.npy` / `.npz` parity and benchmark I/O.
npy = ["dep:ndarray-npy"]
# JSON lines from `bench_backends --json`
bench-json = ["dep:serde_json"]
# Download + vectorize MNIST, Fashion-MNIST, BBC News, 20 Newsgroups.
datasets = ["npy", "dep:ureq", "dep:flate2", "dep:tar", "dep:zip"]
# Optional RLX compute backends (each stands alone; delegate to strict until GPU kernels land).
rlx-cpu = ["cluster"]
rlx-cuda = ["cluster", "dep:rlx", "rlx/cpu", "rlx/cuda"]
rlx-mlx = ["cluster", "dep:rlx", "rlx/cpu", "rlx/mlx"]
rlx-rocm = ["cluster", "dep:rlx", "rlx/cpu", "rlx/rocm"]
rlx-wgpu = ["cluster", "dep:rlx", "rlx/cpu", "rlx/gpu"]
# Enable every RLX backend (optional convenience). Do not name this `rlx` — that
# collides with the implicit feature for optional dep `rlx` and would enable all
# backends whenever any `dep:rlx` backend feature is used.
rlx-all = ["rlx-cpu", "rlx-cuda", "rlx-mlx", "rlx-rocm", "rlx-wgpu"]
# Back-compat alias
rlx_metal = ["rlx-mlx"]

[dev-dependencies]
approx = "0.5"
serde_json = "1"
rand_chacha = "0.3"
criterion = "0.5"
ndarray = { version = "0.16", features = ["rayon"] }
ndarray-npy = "0.9"

[build-dependencies]
cc = "1.2"

[[bin]]
name = "backend_smoke"
path = "src/bin/backend_smoke.rs"
required-features = ["cluster", "npy"]

[[bin]]
name = "bench_backends"
path = "src/bin/bench_backends.rs"
required-features = ["cluster", "npy", "bench-json"]

[[bin]]
name = "bench"
path = "src/bin/bench.rs"
required-features = ["cluster", "npy"]

[[bin]]
name = "bench_huge"
path = "src/bin/bench_huge.rs"
required-features = ["cluster"]

[[bin]]
name = "emb_epoch_diff"
path = "src/bin/emb_epoch_diff.rs"
required-features = ["cluster", "npy"]

[[bin]]
name = "fashion_mnist_fetch"
path = "src/bin/fashion_mnist_fetch.rs"
required-features = ["datasets"]

[[bin]]
name = "mnist_fetch"
path = "src/bin/mnist_fetch.rs"
required-features = ["datasets"]

[[bin]]
name = "mnist_labels"
path = "src/bin/mnist_labels.rs"
required-features = ["cluster", "datasets", "npy"]

[[example]]
name = "bbc_news_clustering"
path = "examples/bbc_news_clustering.rs"
required-features = ["cluster", "datasets"]

[[example]]
name = "cluster_in_memory"
path = "examples/cluster_in_memory.rs"
required-features = ["cluster"]

[[example]]
name = "fashion_mnist_clustering"
path = "examples/fashion_mnist_clustering.rs"
required-features = ["cluster", "datasets"]

[[example]]
name = "news_clustering"
path = "examples/news_clustering.rs"
required-features = ["cluster", "datasets"]

[[example]]
name = "user_clustering"
path = "examples/user_clustering.rs"
required-features = ["cluster", "npy"]

[[bench]]
name = "evoc_bench"
harness = false
required-features = ["cluster", "npy"]