关于keras使用fit_generator中遇到StopIteration

最新推荐文章于 2023-11-04 19:45:19 发布

原创

最新推荐文章于 2023-11-04 19:45:19 发布 · 3.5k 阅读

9 ·

CC 4.0 BY-SA版权

文章标签：

#python #generator #深度学习 #生成器 #tensorflow

本文探讨了在使用Keras的fit_generator训练模型时遇到StopIteration异常的原因及解决办法，通过修改生成器使其无限循环，避免了训练过程中因数据迭代完毕而引发的错误。

Keras中使用fit_generator遇到报错：StopIteration

在keras中训练模型时，训练到一半报错StopIteration

遇到的问题

在这里插入图片描述

_________________________________________________________________
Epoch 1/1000
16107/16107 [==============================] - 15052s 935ms/step - loss: 2.5982 - acc: 0.8210 - val_loss: 2.8054 - val_acc: 0.7029
Epoch 2/1000
    1/16107 [..............................] - ETA: 4:10:08 - loss: 2.5140 - acc: 0.9133Traceback (most recent call last):
  File "train_generator_v1.py", line 104, in <module>
    run()
  File "train_generator_v1.py", line 97, in run
    validation_data=(X, Y), shuffle=True, callbacks=cbks)
  File "/home/limin/.conda/envs/py36-2/lib/python3.6/site-packages/keras/legacy/interfaces.py", line 91, in wrapper
    return func(*args, **kwargs)
  File "/home/limin/.conda/envs/py36-2/lib/python3.6/site-packages/keras/engine/training.py", line 1415, in fit_generator
    initial_epoch=initial_epoch)
  File "/home/limin/.conda/envs/py36-2/lib/python3.6/site-packages/keras/engine/training_generator.py", line 177, in fit_generator
    generator_output = next(output_generator)
  File "/home/limin/.conda/envs/py36-2/lib/python3.6/site-packages/keras/utils/data_utils.py", line 785, in get
    raise StopIteration()
StopIteration

解决办法

给自己自定义的生成器内部嵌套一个while True:

下面是keras官方使用文档中fit_generator的demo

def generate_arrays_from_file(path):
   while

最低0.47元/天解锁文章