Olive/olive/constants.py at 03cee7b998f3185df92eaafc2319abe9dca894e5 · microsoft/Olive · 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
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
# -------------------------------------------------------------------------
# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License.
# --------------------------------------------------------------------------
from enum import IntEnum

from olive.common.config_utils import CaseInsensitiveEnum
from olive.common.utils import StrEnumBase

MSFT_DOMAIN = "com.microsoft"


class Framework(StrEnumBase):
    """Framework of the model."""

    ONNX = "ONNX"
    PYTORCH = "PyTorch"
    QAIRT = "QAIRT"
    QNN = "QNN"
    TENSORFLOW = "TensorFlow"
    OPENVINO = "OpenVINO"


class ModelFileFormat(StrEnumBase):
    """Given a framework, there might be 1 or more on-disk model file format(s), model save/Load logic may differ."""

    ONNX = "ONNX"
    PYTORCH_ENTIRE_MODEL = "PyTorch.EntireModel"
    PYTORCH_STATE_DICT = "PyTorch.StateDict"
    PYTORCH_TORCH_SCRIPT = "PyTorch.TorchScript"
    PYTORCH_SLICE_GPT_MODEL = "PyTorch.SliceGPT"
    TENSORFLOW_PROTOBUF = "TensorFlow.Protobuf"
    TENSORFLOW_SAVED_MODEL = "TensorFlow.SavedModel"
    QAIRT = "QAIRT"
    QAIRT_PREPARED = "QAIRT.Prepared"
    QNN_CPP = "QNN.CPP"
    QNN_LIB = "QNN.LIB"
    QNN_SERIALIZED_BIN = "QNN.SERIALIZED.BIN"
    OPENVINO_IR = "OpenVINO.IR"
    COMPOSITE_MODEL = "Composite"


class Precision(StrEnumBase):
    INT4 = "int4"
    INT8 = "int8"
    INT16 = "int16"
    INT32 = "int32"
    UINT4 = "uint4"
    UINT8 = "uint8"
    UINT16 = "uint16"
    UINT32 = "uint32"
    FP4 = "fp4"
    FP8 = "fp8"
    FP16 = "fp16"
    FP32 = "fp32"
    NF4 = "nf4"
    BF16 = "bf16"


class PrecisionBits(IntEnum):
    BITS2 = 2
    BITS4 = 4
    BITS8 = 8
    BITS16 = 16
    BITS32 = 32


class QuantAlgorithm(CaseInsensitiveEnum):
    AWQ = "awq"
    GPTQ = "gptq"
    HQQ = "hqq"
    RTN = "rtn"
    SPINQUANT = "spinquant"
    QUAROT = "quarot"
    LPBQ = "lpbq"
    SEQMSE = "seqmse"
    ADAROUND = "adaround"


class QuantEncoding(StrEnumBase):
    QDQ = "qdq"
    QOP = "qop"


class DatasetRequirement(StrEnumBase):
    REQUIRED = "dataset_required"
    OPTIONAL = "dataset_optional"
    NOT_REQUIRED = "dataset_not_required"


class OpType(StrEnumBase):
    """Enum for operator types."""

    DequantizeLinear = "DequantizeLinear"
    Gather = "Gather"
    GatherBlockQuantized = "GatherBlockQuantized"
    MatMulNBits = "MatMulNBits"
    MatMul = "MatMul"
    QuickGelu = "QuickGelu"
    Sigmoid = "Sigmoid"
    Mul = "Mul"
    RotaryEmbedding = "RotaryEmbedding"
    Reshape = "Reshape"
    Slice = "Slice"
    Sub = "Sub"
    Add = "Add"
    Concat = "Concat"
    Div = "Div"
    Shape = "Shape"
    Constant = "Constant"
    Custom = "custom"
    PackedAttention = "PackedAttention"
    PackedMultiHeadAttention = "PackedMultiHeadAttention"
    MultiHeadAttention = "MultiHeadAttention"
    Loop = "Loop"


class AccuracyLevel(IntEnum):
    unset = 0
    fp32 = 1
    fp16 = 2
    bf16 = 3
    int8 = 4


class DiffusersModelVariant(StrEnumBase):
    """Diffusion model variants."""

    AUTO = "auto"
    SD = "sd"
    SDXL = "sdxl"
    SD3 = "sd3"
    FLUX = "flux"
    SANA = "sana"


class DiffusersComponent(StrEnumBase):
    """Diffusers pipeline component names."""

    TEXT_ENCODER = "text_encoder"
    TEXT_ENCODER_2 = "text_encoder_2"
    TEXT_ENCODER_3 = "text_encoder_3"
    UNET = "unet"
    TRANSFORMER = "transformer"
    VAE_ENCODER = "vae_encoder"
    VAE_DECODER = "vae_decoder"
    FLUX_TRANSFORMER = "flux_transformer"
    SD3_TRANSFORMER = "sd3_transformer"
    SANA_TRANSFORMER = "sana_transformer"


def precision_bits_from_precision(p):
    mapping = {
        Precision.INT4: PrecisionBits.BITS4,
        Precision.INT8: PrecisionBits.BITS8,
        Precision.INT16: PrecisionBits.BITS16,
        Precision.INT32: PrecisionBits.BITS32,
        Precision.UINT4: PrecisionBits.BITS4,
        Precision.UINT8: PrecisionBits.BITS8,
        Precision.UINT16: PrecisionBits.BITS16,
        Precision.UINT32: PrecisionBits.BITS32,
    }
    return mapping.get(p)