目前,我正在尝试了解TensorFlow中的量化意识培训。我了解,需要伪造的量化节点来收集动态范围信息,以作为量化操作的校准。当我将同一个模型一次比较为“普通” Keras模型,一次比较为量化感知模型时,后者具有更多参数,这很有意义,因为我们需要在量化感知训练期间存储激活的最小值和最大值。
请考虑以下示例:
import tensorflow as tf
from tensorflow.keras import layers
from tensorflow.keras.models import Model
def get_model(in_shape):
inpt = layers.Input(shape=in_shape)
dense1 = layers.Dense(256, activation="relu")(inpt)
dense2 = layers.Dense(128, activation="relu")(dense1)
out = layers.Dense(10, activation="softmax")(dense2)
model = Model(inpt, out)
return model
该模型具有以下摘要:
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_2 (InputLayer) [(None, 784)] 0
_________________________________________________________________
dense_3 (Dense) (None, 256) 200960
_________________________________________________________________
dense_4 (Dense) (None, 128) 32896
_________________________________________________________________
dense_5 (Dense) (None, 10) 1290
=================================================================
Total params: 235,146
Trainable params: 235,146
Non-trainable params: 0
_________________________________________________________________
但是,如果我使我的模型优化知道,它会显示以下摘要:
import tensorflow_model_optimization as tfmot
quantize_model = tfmot.quantization.keras.quantize_model
# q_aware stands for for quantization aware.
q_aware_model = quantize_model(standard)
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~#
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_2 (InputLayer) [(None, 784)] 0
_________________________________________________________________
quantize_layer (QuantizeLaye (None, 784) 3
_________________________________________________________________
quant_dense_3 (QuantizeWrapp (None, 256) 200965
_________________________________________________________________
quant_dense_4 (QuantizeWrapp (None, 128) 32901
_________________________________________________________________
quant_dense_5 (QuantizeWrapp (None, 10) 1295
=================================================================
Total params: 235,164
Trainable params: 235,146
Non-trainable params: 18
_________________________________________________________________
我特别有两个问题:
quantize_layer
的目的是什么?[我感谢任何有助于我(以及其他偶然发现此问题的人)理解量化意识训练的提示或其他材料。
量化层用于将浮点输入转换为int8。量化参数用于输出最小/最大和零点计算。
量化密集层需要一些其他参数:内核的最小/最大和输出激活的最小/最大/零点。