问题一:
我有一个模型需要用到lstm,那请问量化支持序列模型吗?看了开发手册好像没有这方面的介绍。而且官方量化的例子都是图像,没有关于序列的。
问题二:
efficientnet_lite0_config.yaml 里面 ,怎么只有均值,没有方差,图像预处理缩放比例是指像素值除以255吗?
--------------------------------efficientnet_lite0_config.yaml文件的内容------------------------------
# 网络输入的预处理方法,主要有以下几种:
# no_preprocess 不做任何操作
# data_mean 减去通道均值mean_value
# data_scale 对图像像素乘以data_scale系数
# data_mean_and_scale 减去通道均值后再乘以scale系数
# -------------------------------------------------------------------------------------------
# preprocessing methods of network input, available options:
# ‘no_preprocess’ indicates that no preprocess will be made
# ‘data_mean’ indicates that to minus the channel mean, i.e. mean_value
# ‘data_scale’ indicates that image pixels to multiply data_scale ratio
# ‘data_mean_and_scale’ indicates that to multiply scale ratio after channel mean is minused
norm_type: ‘data_mean_and_scale’
# 图像减去的均值, 如果是通道均值,value之间必须用空格分隔
# --------------------------------------------------------------------------
# the mean value minused by image
# note that values must be seperated by space if channel mean value is used
mean_value: 0 0 0
# 图像预处理缩放比例,如果是通道缩放比例,value之间必须用空格分隔
# ---------------------------------------------------------------------------
# scale value of image preprocess
# note that values must be seperated by space if channel scale value is used
scale_value: 0.003921 0.003921 0.003921
问题三:
如果我训练的模型输入有标准化的操作,这里的内容也是要该的吗?
--------------------------------preprocess.py文件的内容------------------------------
def calibration_transformers():
“”"
step:
1、crop size 224 * 224 from padded center
2、resize to 224 * 224
3、Scale from 0 ~ 1 to 0 ~ 255
“”"
transformers = [
PaddedCenterCropTransformer(image_size=224, crop_pad=32),
# Cubic interpolation resize
ResizeTransformer(target_size=(224, 224), mode=‘skimage’, method=3),
ScaleTransformer(scale_value=255),
]
return transformers