Python TensorFlow键错误:';RegexReplace';使用tf.contrib.predictor.from_saved_模型对DNNClassifier进行预测时
我试图恢复一个估计器并对其进行预测 系统配置: CentOS 7-x64-CPU tensorflow==1.5 这就是我如何保存估算器DNNClassifier的方法Python TensorFlow键错误:';RegexReplace';使用tf.contrib.predictor.from_saved_模型对DNNClassifier进行预测时,python,tensorflow,centos7,Python,Tensorflow,Centos7,我试图恢复一个估计器并对其进行预测 系统配置: CentOS 7-x64-CPU tensorflow==1.5 这就是我如何保存估算器DNNClassifier的方法 def serving_input_receiver_fn(): inputs = {"embeddings": tf.placeholder(shape=[None], dtype=tf.string)} return tf.estimator.export.ServingInputReceiver(
def serving_input_receiver_fn():
inputs = {"embeddings": tf.placeholder(shape=[None], dtype=tf.string)}
return tf.estimator.export.ServingInputReceiver(inputs, inputs)
export_dir = estimator.export_savedmodel(
export_dir_base='models/run7',
serving_input_receiver_fn=serving_input_receiver_fn)
我使用tf.contrib.predictor.from\u saved\u model重新加载了它,如下所示
estimator = tf.contrib.predictor.from_saved_model('model/1528805269/')
但是,我在上面一行中得到了错误KeyError:'RegexReplace'
完全错误:
Traceback (most recent call last):
File "app.py", line 38, in <module>
load_model()
File "app.py", line 21, in load_model
estimator = tf.contrib.predictor.from_saved_model('model/1528805269/')
File "/var/www/html/my_project/prj_v4/lib/python3.6/site-packages/tensorflow/contrib/predictor/predictor_factories.py", line 129, in from_saved_model
graph=graph)
File "/var/www/html/my_project/prj_v4/lib/python3.6/site-packages/tensorflow/contrib/predictor/saved_model_predictor.py", line 156, in __init__
loader.load(self._session, tags.split(','), export_dir)
File "/var/www/html/my_project/prj_v4/lib/python3.6/site-packages/tensorflow/python/saved_model/loader_impl.py", line 216, in load
saver = tf_saver.import_meta_graph(meta_graph_def_to_load, **saver_kwargs)
File "/var/www/html/my_project/prj_v4/lib/python3.6/site-packages/tensorflow/python/training/saver.py", line 1838, in import_meta_graph
**kwargs)
File "/var/www/html/my_project/prj_v4/lib/python3.6/site-packages/tensorflow/python/framework/meta_graph.py", line 660, in import_scoped_meta_graph
producer_op_list=producer_op_list)
File "/var/www/html/my_project/prj_v4/lib/python3.6/site-packages/tensorflow/python/util/deprecation.py", line 316, in new_func
return func(*args, **kwargs)
File "/var/www/html/my_project/prj_v4/lib/python3.6/site-packages/tensorflow/python/framework/importer.py", line 433, in import_graph_def
_RemoveDefaultAttrs(op_dict, producer_op_list, graph_def)
File "/var/www/html/my_project/prj_v4/lib/python3.6/site-packages/tensorflow/python/framework/importer.py", line 211, in _RemoveDefaultAttrs
op_def = op_dict[node.op]
KeyError: 'RegexReplace'
回溯(最近一次呼叫最后一次):
文件“app.py”,第38行,在
负载_模型()
文件“app.py”,第21行,在load_模型中
估计器=tf.contrib.predictor.from_saved_模型('model/1528805269/'))
文件“/var/www/html/my_project/prj_v4/lib/python3.6/site packages/tensorflow/contrib/predictor/predictor_factories.py”,第129行,来自保存的模型
图=图)
文件“/var/www/html/my_project/prj_v4/lib/python3.6/site packages/tensorflow/contrib/predictor/saved_model_predictor.py”,第156行,在__
loader.load(self.\u会话,tags.split(','),export\u dir)
文件“/var/www/html/my_project/prj_v4/lib/python3.6/site packages/tensorflow/python/saved_model/loader_impl.py”,第216行,在load中
saver=tf_saver.import_meta_graph(meta_graph_def_to_load,**saver_kwargs)
文件“/var/www/html/my_project/prj_v4/lib/python3.6/site packages/tensorflow/python/training/saver.py”,第1838行,在导入元图中
**kwargs)
文件“/var/www/html/my_project/prj_v4/lib/python3.6/site packages/tensorflow/python/framework/meta_graph.py”,第660行,在导入范围的meta_图中
制片人名单=制片人名单)
文件“/var/www/html/my_project/prj_v4/lib/python3.6/site packages/tensorflow/python/util/deprecation.py”,第316行,新函数
返回函数(*args,**kwargs)
文件“/var/www/html/my_project/prj_v4/lib/python3.6/site packages/tensorflow/python/framework/importer.py”,第433行,在import_graph_def中
_移除的默认属性(操作指令、生产者操作列表、图形定义)
文件“/var/www/html/my_project/prj_v4/lib/python3.6/site packages/tensorflow/python/framework/importer.py”,第211行,位于RemoveDefaultAttrs中
op_def=op_dict[node.op]
KeyError:'RegexReplace'
有什么建议可能出了什么问题 1.5版的TensorFlow中没有RegexReplace,所以我猜您正在1.6+中保存SavedModel,并在1.5版中使用它(1.5版中不存在op,因此出现错误)。啊!。。。好吧,我想我有麻烦了。。。。我使用tensorflow hub的一些模块在google colab上训练了一个模型,需要tf 1.7>=,我的系统存在非法指令核心转储问题,不允许tf版本>1.5。。。我在tf 1.8中训练了一些模型,并在1.5中运行,没有任何问题…我想这一次它不会工作…感谢您的响应…对,AVX指令用于预打包的版本。您还可以从没有AVX的源代码构建更新的版本。