LLamaSharp/LLama/Native/LLamaModelQuantizeParams.cs at 39457059bcadac0fe0c64012fe4d0f54394fbe69 · martindevans/LLamaSharp · GitHub

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
using System;

namespace LLama.Native
{
    /// <summary>
    /// Quantizer parameters used in the native API
    /// </summary>
    /// <remarks>llama_model_quantize_params</remarks>
    [StructLayout(LayoutKind.Sequential)]
    public struct LLamaModelQuantizeParams
    {
        /// <summary>
        /// number of threads to use for quantizing, if &lt;=0 will use std::thread::hardware_concurrency()
        /// </summary>
        public int nthread;

        /// <summary>
        /// quantize to this llama_ftype
        /// </summary>
        public LLamaFtype ftype;

        /// <summary>
        /// output tensor type
        /// </summary>
        public GGMLType output_tensor_type;

        /// <summary>
        /// token embeddings tensor type
        /// </summary>
        public GGMLType token_embedding_type;

        /// <summary>
        /// allow quantizing non-f32/f16 tensors
        /// </summary>
        public bool allow_requantize
        {
            get => Convert.ToBoolean(_allow_requantize);
            set => _allow_requantize = Convert.ToSByte(value);
        }
        private sbyte _allow_requantize;

        /// <summary>
        /// quantize output.weight
        /// </summary>
        public bool quantize_output_tensor
        {
            get => Convert.ToBoolean(_quantize_output_tensor);
            set => _quantize_output_tensor = Convert.ToSByte(value);
        }
        private sbyte _quantize_output_tensor;

        /// <summary>
        /// only copy tensors - ftype, allow_requantize and quantize_output_tensor are ignored
        /// </summary>
        public bool only_copy
        {
            get => Convert.ToBoolean(_only_copy);
            set => _only_copy = Convert.ToSByte(value);
        }
        private sbyte _only_copy;

        /// <summary>
        /// quantize all tensors to the default type
        /// </summary>
        public bool pure
        {
            get => Convert.ToBoolean(_pure);
            set => _pure = Convert.ToSByte(value);
        }
        private sbyte _pure;

        /// <summary>
        /// quantize to the same number of shards
        /// </summary>
        public bool keep_split
        {
            get => Convert.ToBoolean(_keep_split);
            set => _keep_split = Convert.ToSByte(value);
        }
        private sbyte _keep_split;

        /// <summary>
        /// pointer to importance matrix data
        /// </summary>
        public IntPtr imatrix;

        /// <summary>
        /// pointer to vector containing overrides
        /// </summary>
        public IntPtr kv_overrides;

        /// <summary>
        /// Create a LLamaModelQuantizeParams with default values
        /// </summary>
        /// <returns></returns>
        public static LLamaModelQuantizeParams Default()
        {
            return llama_model_quantize_default_params();

            [DllImport(NativeApi.libraryName, CallingConvention = CallingConvention.Cdecl)]
            static extern LLamaModelQuantizeParams llama_model_quantize_default_params();
        }
    }
}