ferrum-kernels 0.7.2

Unified compute kernels (CUDA/Metal/CPU) and model runner for Ferrum inference
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

# THIS FILE IS AUTOMATICALLY GENERATED BY CARGO
#
# When uploading crates to the registry Cargo will automatically
# "normalize" Cargo.toml files for maximal compatibility
# with all versions of Cargo and also rewrite `path` dependencies
# to registry (e.g., crates.io) dependencies.
#
# If you are reading this file be aware that the original Cargo.toml
# will likely look very different (and much more reasonable).
# See Cargo.toml.orig for the original contents.

[package]
edition = "2021"
name = "ferrum-kernels"
version = "0.7.2"
authors = ["Ferrum Team"]
build = "build.rs"
autolib = false
autobins = false
autoexamples = false
autotests = false
autobenches = false
description = "Unified compute kernels (CUDA/Metal/CPU) and model runner for Ferrum inference"
readme = "README.md"
license = "MIT"

[features]
cuda = [
    "candle-core/cuda",
    "dep:cudarc",
    "marlin",
    "cudarc?/nccl",
]
default = []
marlin = []
metal = [
    "dep:metal",
    "dep:ferrum-attention",
    "ferrum-attention/metal",
]
tensor-parallel = ["cuda"]
triton-kernels = ["cuda"]

[lib]
name = "ferrum_kernels"
path = "src/lib.rs"

[[test]]
name = "cudarc_graph_repro"
path = "tests/cudarc_graph_repro.rs"

[[test]]
name = "triton_add_bias_eq"
path = "tests/triton_add_bias_eq.rs"

[[test]]
name = "triton_fused_add_rms_norm_eq"
path = "tests/triton_fused_add_rms_norm_eq.rs"

[[test]]
name = "triton_fused_silu_mul_eq"
path = "tests/triton_fused_silu_mul_eq.rs"

[[test]]
name = "triton_gelu_eq"
path = "tests/triton_gelu_eq.rs"

[[test]]
name = "triton_layer_norm_eq"
path = "tests/triton_layer_norm_eq.rs"

[[test]]
name = "triton_residual_add_eq"
path = "tests/triton_residual_add_eq.rs"

[[test]]
name = "triton_residual_add_inplace_eq"
path = "tests/triton_residual_add_inplace_eq.rs"

[[test]]
name = "triton_rms_norm_eq"
path = "tests/triton_rms_norm_eq.rs"

[[test]]
name = "triton_softmax_eq"
path = "tests/triton_softmax_eq.rs"

[[test]]
name = "triton_w4a16_eq"
path = "tests/triton_w4a16_eq.rs"

[dependencies.candle-core]
version = "0.9.2"

[dependencies.cudarc]
version = "0.19"
features = [
    "driver",
    "cublas",
    "f16",
    "nvrtc",
]
optional = true
default-features = false

[dependencies.ferrum-attention]
version = "0.7.0"
optional = true

[dependencies.ferrum-types]
version = "0.7.2"

[dependencies.half]
version = "2.5.0"
features = [
    "num-traits",
    "use-intrinsics",
    "rand_distr",
]

[dependencies.metal]
version = "0.31"
optional = true

[dependencies.tracing]
version = "0.1"

[build-dependencies.bindgen_cuda]
version = "0.1.6"