LiteRT 8 位量化规范

以下文档概述了 LiteRT 的 8 位量化方案规范。本指南旨在帮助硬件开发者为使用量化 LiteRT 模型进行推理提供硬件支持。

规范摘要

我们提供的是规范，只有遵循该规范，我们才能保证行为。我们还了解到，不同的硬件可能具有偏好和限制，这可能会导致在实现规范时出现细微偏差，从而导致实现不是位精确的。虽然在大多数情况下，这可能是可以接受的（我们将提供一系列测试，据我们所知，这些测试包括我们从多个模型中收集的每项操作容差），但机器学习（以及最常见的深度学习）的性质使得我们无法提供任何硬性保证。

8 位量化使用以下公式来近似浮点值。

\[real\_value = (int8\_value - zero\_point) \times scale\]

每个轴（即卷积运算中的每个通道）或每个张量的权重由 int8 二进制补码值表示，范围为 [-127, 127]，零点等于 0。每个张量的激活/输入由 int8 二进制补码值表示，范围为 [-128, 127]，零点范围为 [-128, 127]。

下文介绍了特定操作的其他例外情况。

有符号整数与无符号整数

LiteRT 量化将主要优先考虑 8 位量化的工具和内核。int8这是为了方便将零点等于 0 的对称量化表示出来。此外，许多后端还针对 int8xint8 累积进行了额外的优化。

按轴与按张量

按张量量化是指每个张量将有一个比例和/或零点。按轴量化是指 quantized_dimension 中的每个切片都将有一个缩放比例和/或 zero_point。量化维度用于指定 Tensor 形状中与缩放比例和零点对应的维度。例如，张量 t，其中 dims=[4, 3, 2, 1] 具有量化参数：scale=[1.0, 2.0, 3.0]、zero_point=[1, 2, 3]、quantization_dimension=1，将沿 t 的第二个维度进行量化：

t[:, 0, :, :] will have scale[0]=1.0, zero_point[0]=1
t[:, 1, :, :] will have scale[1]=2.0, zero_point[1]=2
t[:, 2, :, :] will have scale[2]=3.0, zero_point[2]=3

通常，quantized_dimension 是卷积权重的 output_channel，但从理论上讲，它可以是与内核实现中的每个点积相对应的维度，从而在不影响性能的情况下实现更精细的量化。这在很大程度上提高了准确性。

TFLite 针对越来越多的操作提供了按轴支持。截至本文档发布时，Conv2d 和 DepthwiseConv2d 已获得支持。

对称与非对称

激活是不对称的：它们的零点可以位于有符号 int8 范围 [-128, 127] 内的任何位置。许多激活函数本质上是不对称的，而零点是一种相对便宜的方式，可有效获得额外的二进制精度。由于激活仅乘以恒定权重，因此可以对恒定零点值进行大量优化。

权重是对称的：强制零点等于 0。权重值会乘以动态输入和激活值。这意味着，将权重零点与激活值相乘会产生不可避免的运行时成本。通过强制将零点设为 0，我们可以避免此成本。

数学解释：这与 arXiv:1712.05877 中的第 2.3 节类似，不同之处在于我们允许缩放值按轴变化。这可以轻松推广，如下所示：

$A$ 是一个 $m \times n$ 的量化激活矩阵。
$B$ 是一个 $n \times p$ 量化权重矩阵。
假设将长度均为 $n$的 $A$ 的第 $j$ 行 $a_j$与 $B$ 的第 $k$ 列 $b_k$ 相乘。量化后的整数值和零点值分别为 $q_a$、$z_a$ 和 $q_b$、$z_b$。

\[a_j \cdot b_k = \sum_{i=0}^{n} a_{j}^{(i)} b_{k}^{(i)} = \sum_{i=0}^{n} (q_{a}^{(i)} - z_a) (q_{b}^{(i)} - z_b) = \sum_{i=0}^{n} q_{a}^{(i)} q_{b}^{(i)} - \sum_{i=0}^{n} q_{a}^{(i)} z_b - \sum_{i=0}^{n} q_{b}^{(i)} z_a + \sum_{i=0}^{n} z_a z_b\]

$\sum_{i=0}^{n} q_{a}^{(i)} q_{b}^{(i)}$ 项是不可避免的，因为它是输入值和权重值的点积。

$\sum_{i=0}^{n} q_{b}^{(i)} z_a$ 和 $\sum_{i=0}^{n} z_a z_b$ 项由在每次推理调用中保持不变的常量组成，因此可以预先计算。

由于每次推理时激活都会发生变化，因此需要在每次推理时计算 $\sum_{i=0}^{n} q_{a}^{(i)} z_b$ 项。通过强制使权重保持对称，我们可以消除此项的成本。

int8 量化运算符规范

下面介绍了 int8 tflite 内核的量化要求：

ADD
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

AVERAGE_POOL_2D
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

CONCATENATION
  Input ...:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

CONV_2D
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1 (Weight):
    data_type  : int8
    range      : [-127, 127]
    granularity: per-axis (dim = 0)
    restriction: zero_point = 0
  Input 2 (Bias):
    data_type  : int32
    range      : [int32_min, int32_max]
    granularity: per-axis
    restriction: (scale, zero_point) = (input0_scale * input1_scale[...], 0)
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

DEPTHWISE_CONV_2D
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1 (Weight):
    data_type  : int8
    range      : [-127, 127]
    granularity: per-axis (dim = 3)
    restriction: zero_point = 0
  Input 2 (Bias):
    data_type  : int32
    range      : [int32_min, int32_max]
    granularity: per-axis
    restriction: (scale, zero_point) = (input0_scale * input1_scale[...], 0)
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

FULLY_CONNECTED
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1 (Weight):
    data_type  : int8
    range      : [-127, 127]
    granularity: per-axis (dim = 0)
    restriction: zero_point = 0
  Input 2 (Bias):
    data_type  : int32
    range      : [int32_min, int32_max]
    granularity: per-tensor
    restriction: (scale, zero_point) = (input0_scale * input1_scale[...], 0)
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

L2_NORMALIZATION
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
    restriction: (scale, zero_point) = (1.0 / 128.0, 0)

LOGISTIC
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
    restriction: (scale, zero_point) = (1.0 / 256.0, -128)

MAX_POOL_2D
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

MUL
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

RESHAPE
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

RESIZE_BILINEAR
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

SOFTMAX
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
    restriction: (scale, zero_point) = (1.0 / 256.0, -128)

SPACE_TO_DEPTH
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

TANH
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
    restriction: (scale, zero_point) = (1.0 / 128.0, 0)

PAD
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

GATHER
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

BATCH_TO_SPACE_ND
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

SPACE_TO_BATCH_ND
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

TRANSPOSE
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

MEAN
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

SUB
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

SUM
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

SQUEEZE
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

LOG_SOFTMAX
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
    restriction: (scale, zero_point) = (16.0 / 256.0, 127)

MAXIMUM
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

ARG_MAX
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

MINIMUM
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

LESS
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

PADV2
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

GREATER
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

GREATER_EQUAL
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

LESS_EQUAL
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

SLICE
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  restriction: Input and outputs must all have same scale/zero_point

EQUAL
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

NOT_EQUAL
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Input 1:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

SHAPE
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

QUANTIZE (Requantization)
  Input 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor
  Output 0:
    data_type  : int8
    range      : [-128, 127]
    granularity: per-tensor

参考

arXiv:1712.05877