gllm-kernels 0.1.3

Low-level attention kernels for gllm with CUDA/ROCm support
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

[[bench]]
harness = false
name = "attention_bench"
path = "benches/attention_bench.rs"

[dependencies.burn]
default-features = false
version = "0.20.0-pre.6"

[dependencies.burn-cuda]
optional = true
version = "0.20.0-pre.6"

[dependencies.burn-fusion]
optional = true
version = "0.20.0-pre.6"

[dependencies.burn-ndarray]
optional = true
version = "0.20.0-pre.6"

[dependencies.burn-wgpu]
optional = true
version = "0.20.0-pre.6"

[dependencies.crossbeam-channel]
version = "0.5"

[dependencies.cudarc]
default-features = false
features = ["std", "driver", "nvrtc", "f16", "cuda-12090", "dynamic-loading"]
optional = true
version = "0.18"

[dependencies.half]
optional = true
version = "2.7"

[dependencies.log]
version = "0.4"

[dev-dependencies.criterion]
version = "0.5"

[features]
cpu = ["burn-ndarray"]
cuda = ["burn-cuda"]
cuda-kernel = ["cudarc", "half"]
default = ["cpu"]
flash-attention-v3 = ["flash-attention-v3-wgmma", "flash-attention-v3-async", "flash-attention-v3-fp8", "flash-attention-v3-block-quant"]
flash-attention-v3-async = []
flash-attention-v3-block-quant = []
flash-attention-v3-fp8 = []
flash-attention-v3-wgmma = []
fusion = ["burn/fusion", "burn-fusion"]
nccl = ["cudarc/nccl", "cuda"]
rocm-kernel = ["half"]
wgpu = ["burn-wgpu"]

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

[package]
authors = ["gllm contributors"]
autobenches = false
autobins = false
autoexamples = false
autolib = false
autotests = false
build = false
categories = ["algorithms", "science"]
description = "Low-level attention kernels for gllm with CUDA/ROCm support"
edition = "2021"
homepage = "https://github.com/putao520/gllm-kernels"
keywords = ["attention", "kernels", "burn", "gpu", "tensor"]
license = "Apache-2.0"
name = "gllm-kernels"
readme = "README.md"
repository = "https://github.com/putao520/gllm-kernels"
version = "0.1.3"

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

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

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

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