converter = tf.lite.TocoConverter.from_keras_model_file("model2.h5")
tflite_quantized_model=converter.convert()
open("quantized_model.tflite", "wb").write(tflite_quantized_model)
找不到lite,所以tensorflow-gpu=1.8 不行啦
升级到1.13.cudu和cudnn不行
降级到1.12,如下的方案:
https://stackoverflow.com/questions/53914785/how-to-fix-this-module-tensorflow-has-no-attribute-lite
converter = tf.contrib.lite.TFLiteConverter.from_keras_model_file("sample_android.h5")
tflite_model = converter.convert()
open("quantized_model.tflite", "wb").write(tflite_quantized_model)
https://stackoverflow.com/questions/53680660/attributeerror-module-tensorflow-has-no-attribute-lite-in-keras-model-to-te
In tensorflow 1.12 it should be converter = tf.contrib.lite.TFLiteConverter.from_keras_model_file(keras_file)
还是不行。
果断用tensorflow算啦。
import tensorflow as tf converter = tf.lite.TFLiteConverter.from_keras_model_file("live_dete_generator_validation_data.h5") # converter = tf.contrib.lite.TFLiteConverter.from_keras_model_file("live_dete_generator_validation_data.h5") tflite_quantized_model= converter.convert() open("quantized_model.tflite", "wb").write(tflite_quantized_model)
还是不行啦。。。。