时间:2021-05-22
我就废话不多说了,大家还是直接看代码吧~
</pre><pre code_snippet_id="1947416" snippet_file_name="blog_20161025_1_3331239" name="code" class="python">
# coding:utf-8"""If you want to load pre-trained weights that include convolutions (layers Convolution2D or Convolution1D),be mindful of this: Theano and TensorFlow implement convolution in different ways (TensorFlow actually implements correlation, much like Caffe),and thus, convolution kernels trained with Theano (resp. TensorFlow) need to be converted before being with TensorFlow (resp. Theano)."""from keras import backend as Kfrom keras.utils.np_utils import convert_kernelfrom text_classifier import keras_text_classifierimport sys def th2tf( model): import tensorflow as tf ops = [] for layer in model.layers: if layer.__class__.__name__ in ['Convolution1D', 'Convolution2D']: original_w = K.get_value(layer.W) converted_w = convert_kernel(original_w) ops.append(tf.assign(layer.W, converted_w).op) K.get_session().run(ops) return model def tf2th(model): for layer in model.layers: if layer.__class__.__name__ in ['Convolution1D', 'Convolution2D']: original_w = K.get_value(layer.W) converted_w = convert_kernel(original_w) K.set_value(layer.W, converted_w) return model def conv_layer_converted(tf_weights, th_weights, m = 0): """ :param tf_weights: :param th_weights: :param m: 0-tf2th, 1-th2tf :return: """ if m == 0: # tf2th tc = keras_text_classifier(weights_path=tf_weights) model = tc.loadmodel() model = tf2th(model) model.save_weights(th_weights) elif m == 1: # th2tf tc = keras_text_classifier(weights_path=th_weights) model = tc.loadmodel() model = th2tf(model) model.save_weights(tf_weights) else: print("0-tf2th, 1-th2tf") returnif __name__ == '__main__': if len(sys.argv) < 4: print("python tf_weights th_weights <0|1>\n0-tensorflow to theano\n1-theano to tensorflow") sys.exit(0) tf_weights = sys.argv[1] th_weights = sys.argv[2] m = int(sys.argv[3]) conv_layer_converted(tf_weights, th_weights, m)补充知识:keras学习之修改底层为TensorFlow还是theano
我们知道,keras的底层是TensorFlow或者theano
要知道我们是用的哪个为底层,只需要import keras即可显示
修改方法:
打开
修改
以上这篇keras实现theano和tensorflow训练的模型相互转换就是小编分享给大家的全部内容了,希望能给大家一个参考,也希望大家多多支持。
声明:本页内容来源网络,仅供用户参考;我单位不保证亦不表示资料全面及准确无误,也不保证亦不表示这些资料为最新信息,如因任何原因,本网内容或者用户因倚赖本网内容造成任何损失或损害,我单位将不会负任何法律责任。如涉及版权问题,请提交至online#300.cn邮箱联系删除。
Keras的底层库使用Theano或TensorFlow,这两个库也称为Keras的后端。无论是Theano还是TensorFlow,都需要提前定义好网络的结构
众所周知tensorflow造势虽大却很难用,因此推荐使用Keras,它缺省是基于tensorflow的,但通过修改keras.json也可以用于theano。
实验室新装了keras,发现keras默认后端是tensorflow,想换回theano,看了官方文档也没搞懂,最终搞定,很简单。中文文档的描述:keras中文
tensorflow中如果要对神经网络模型进行训练,需要把训练数据转换为tfrecord格式才能被读取,tensorflow的model文件里直接提供了相应的脚
目标是想把在服务器上用pytorch训练好的模型转换为可以在移动端运行的tflite模型。最直接的思路是想把pytorch模型转换为tensorflow的模型,