[Question] labeling模型进行推理时随次数的增加而进程占用的内存逐渐增多，是否存在内存内存泄漏现象？

[mejackomg]

kashgari 1.1.5，tf1.15，我的代码：
import tensorflow as tf
import kashgari
from tensorflow.python.keras.backend import set_session, clear_session

由于需要使用多个模型，对每个模型定义了全局graph和session

载入checkpoint文件

graph = tf.Graph()
tf_global_config = tf.ConfigProto(inter_op_parallelism_threads=1, intra_op_parallelism_threads=1)
sess = tf.Session(config=tf_global_config, graph=graph)
with graph.as_default():
set_session(sess)
model = kashgari.utils.load_model("./rbt6_bigru_crf/", load_weights=False)
model.tf_model.load_weights("./rbt6_bigru_crf/saved-model-53-0.398.hdf5")

推理

with graph.as_default():
set_session(sess)
model.predict(tokens)

在stackflow上看到有说是因为推理时tf往graph中新增了op导致的，在载入模型后使用graph.finalize()确实会报错：
File "/Users/caijing/opt/anaconda3/envs/nlp_embedding/lib/python3.6/site-packages/tensorflow_core/python/framework/ops.py", line 2998, in _check_not_finalized
raise RuntimeError("Graph is finalized and cannot be modified.")

请问大家遇到过这种内存泄漏的情况吗?
应用环境不能用tf_serving，所以只能采用这种方式进行在线推理。

[mejackomg]

看上去像是keras的锅：keras-team/keras#13118

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.