llama-gguf 0.14.0

A high-performance Rust implementation of llama.cpp - LLM inference engine with full GGUF 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
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
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173

#[cfg(feature = "hailo")]
pub(crate) mod context;
pub mod config;
#[cfg(feature = "hailo")]
pub(crate) mod compiler;
#[cfg(feature = "hailo")]
pub mod gpu_only;

use crate::backend::{Backend, BackendError, BackendResult};
use crate::tensor::{DType, Tensor};

pub use config::{HailoConfig, HailoQuantization, HefManifest};

#[cfg(feature = "hailo")]
pub use context::{check_device_available, HailoContext};
#[cfg(feature = "hailo")]
pub use compiler::{check_cache, model_config_hash, resolve_hef_path};

pub struct HailoBackend {
    _config: HailoConfig,
}

impl HailoBackend {
    pub fn new() -> Result<Self, BackendError> {
        Self::with_config(HailoConfig::default())
    }

    #[cfg(feature = "hailo")]
    pub fn with_config(config: HailoConfig) -> Result<Self, BackendError> {
        context::check_device_available()?;
        Ok(Self { _config: config })
    }

    #[cfg(not(feature = "hailo"))]
    pub fn with_config(_config: HailoConfig) -> Result<Self, BackendError> {
        Err(BackendError::NotAvailable(
            "Hailo support not compiled. Build with --features hailo".to_string(),
        ))
    }
}

impl Backend for HailoBackend {
    fn name(&self) -> &str {
        "hailo"
    }

    fn is_available(&self) -> bool {
        false
    }

    fn alloc(&self, _shape: &[usize], _dtype: DType) -> BackendResult<Tensor> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn copy_to(&self, _tensor: &Tensor) -> BackendResult<Tensor> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn add(&self, _a: &Tensor, _b: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn mul(&self, _a: &Tensor, _b: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn scale(&self, _a: &Tensor, _scalar: f32, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn silu(&self, _x: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn gelu(&self, _x: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn softmax(&self, _x: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn rms_norm(
        &self,
        _x: &Tensor,
        _weight: &Tensor,
        _eps: f32,
        _out: &mut Tensor,
    ) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn matmul(&self, _a: &Tensor, _b: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn matvec(&self, _a: &Tensor, _b: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn vec_mat(&self, _a: &Tensor, _b: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn dequantize(&self, _src: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn matvec_q(&self, _a: &Tensor, _b: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn vec_mat_q(&self, _a: &Tensor, _b: &Tensor, _out: &mut Tensor) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn rope(
        &self,
        _q: &mut Tensor,
        _k: &mut Tensor,
        _pos: usize,
        _freq_base: f32,
        _freq_scale: f32,
        _use_neox: bool,
    ) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }

    fn attention(
        &self,
        _q: &Tensor,
        _k: &Tensor,
        _v: &Tensor,
        _out: &mut Tensor,
        _scale: f32,
    ) -> BackendResult<()> {
        Err(BackendError::Unsupported(
            "Hailo does not support individual tensor ops".into(),
        ))
    }
}