LiteRT 8 位元量化規格

以下文件概略說明 LiteRT 的 8 位元規格量化架構此協議旨在協助硬體開發人員硬體支援，透過量化 LiteRT 模型進行推論。

規格摘要

我們提供了規格，但目前我們只能針對達到指定標準時的行為我們也瞭解不同的硬體設置偏好設定和限制，可能會導致但會使實作的規格不精確。大多數的情況下可能可以接受 (不過，我們會提供一系列最有利於 Google Cloud 的個別作業容忍度測試也就是從多個模型收集而來，代表機器學習 (和深度學習) 的本質因此無法提供任何硬性保證。

8 位元量化會使用下列指令，近似浮點值。

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

每個軸 (又稱「Conv Ops」中，每個管道) 或每個張量權重的 int8兩個在 [-127, 127] 範圍內相等的值 (零點等於) 設為 0。個別張量啟用/輸入以 int8 兩項的補充表示範圍內 [-128, 127] 的值，其中 [-128, 127] 的範圍為零。

其他例外情況則在下文中說明。

帶正負號整數與無正負號整數的差異

LiteRT 量化功能主要會優先處理 int8 量化 (適用於 8 位元)。以便對稱零點等於 0 的量化還有許多後端會針對「int8xint8」累計進行額外的最佳化調整。

每軸與每張量

每張張量量化是指每個整個張量每個軸的量化代表只有一個量表和/或 quantized_dimension 中每個配量 zero_point。量化維度會指定張量縮放和零點的維度尺寸物件例如具有 dims=[4, 3, 2, 1] 的張量 t 量化參數：scale=[1.0, 2.0, 3.0]、zero_point=[1, 2, 3]、 quantization_dimension=1會因 t 的第二個維度進行量化：

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 是權重的 output_channel 但理論上可是與各個符記對應的維度核心實作中的 dot-product，可以提高量化精細程度而不會影響效能這個做法可大幅提升準確率。

TFLite 提供的作業數量與日俱增。於本文提供 Conv2d 和 DepthwiseConv2d 的支援。

對稱與非對稱式

啟動作業不對稱，它們可以在已簽署 int8，範圍為 [-128, 127]。許多啟動方式本身不具對稱性，零點是一種相對便宜的方法二元位數的精確度因為啟動只會乘以常數常數的零點值則可極度最佳化

權重為對稱式：強制使零點等於 0。權重值為乘以動態輸入值和啟用值也就是說難以避免的執行階段成本，也就是將權重的零點乘以啟用值。強制執行零點為 0 就能避免成本。

數學說明：這與第 2.3 節內容相似 arXiv:1712.05877，但差異除外我們允許每個軸的比例值這會顯而易見如下：

$A$ 是 $m\times n$ 量化啟動矩陣。
$B$ 是量化權重的 $n\times p$ 矩陣。
建議將 $A$的第 $j$ 列，$a_j$ 乘以 $k$欄 $B$、$b_k$，兩者的長度都是 $n$。量化整數值零點值分別為 $q_a$、$z_a$ 和 $q_b$、$z_b$。

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

這個字詞 $\sum_{i=0}^{n} q_{a}^{(i)} q_{b}^{(i)}$ 不可避免，因為計算輸入值和重量值的內積。

條款 $\sum_{i=0}^{n} q_{b}^{(i)} z_a$ 與 $\sum_{i=0}^{n} z_a z_b$ 條款為是由常數構成，這些常數在每次推論叫用時保持不變預先計算

每次推論都需要計算 $\sum_{i=0}^{n} q_{a}^{(i)} z_b$ 字詞因為啟用會變更每次推論強制要求權重我們可以移除這個字詞的費用

int8 量化運算子規格

以下將說明 int8 tflite 核心的量化需求：

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

參考資料

arXiv:1712.05877