tensorflow中checkpoint断点生成，保存，下载参数续训。

最新推荐文章于 2022-07-08 21:13:53 发布

原创

最新推荐文章于 2022-07-08 21:13:53 发布 · 1.3k 阅读

6 ·

CC 4.0 BY-SA版权

文章标签：

#tensorflow #深度学习

本文介绍如何使用四段简单的TensorFlow代码实现模型参数的保存、下载及续训练过程。首先设定检查点文件路径，接着通过判断索引文件的存在来决定是否加载已保存的模型权重。然后定义回调函数以便在训练过程中自动保存模型参数。

四段代码实现参数保存、下载、续训练

1.设置ckpt文件保存路径

其中保存有模型参数（特定文件类型


checkpoint_save_path = "checkpoint/mnist.ckpt"

2.判断是否有索引，如果保存了参数模型，会有相应索引文件

if os.path.exists(checkpoint_save_path + '.index'):
    print("---------------loading the model---------------")
    model.load_weights(checkpoint_save_path)

3.callback定义模型参数保存相关内容

cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_save_path,
                                                 # save_best_only=True,
                                                 save_weights_only=

最低0.47元/天解锁文章

5 条评论

匹数 2024.08.03
您好，报错这样ValueError: The filepath provided must end in `.keras` (Keras model format). Received: filepath=checkpoint/mnist.ckpt，我看官网tensorflow2.16也有ckpt格式，但是报错了。
- 匹数回复匹数 2024.08.03
  好像新版不支持ckpt了， filepath的格式：string or PathLike, path to save the model file. filepath can contain named formatting options, which will be filled the value of epoch and keys in logs (passed in on_epoch_end). The filepath name needs to end with ".weights.h5" when save_weights_only=True or should end with ".keras" when checkpoint saving the whole model (default). For example: if filepath is "{epoch:02d}-{val_loss:.2f}.keras", then the model checkpoints will be saved with the epoch number and the validation loss in the filename. The directory of the filepath should not be reused by any other callbacks to avoid conflicts.

quanlaibahua 2023.05.25
你好，我想问一下我的代码加上断电续训就是这样了，找了好久问题，能解惑吗 Traceback (most recent call last): File "D:\Python tf练习\class4\FASHION_FC\p16_fashion_train_ex3.py", line 27, in <module> history = model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1, File "D:\python tf\envs\TF2.1\lib\site-packages\keras\utils\traceback_utils.py", line 70, in error_handler raise e.with_traceback(filtered_tb) from None File "D:\python tf\envs\TF2.1\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 513, in recursive_create_dir_v2 _pywrap_file_io.RecursivelyCreateDir(compat.path_to_bytes(path)) tensorflow.python.framework.errors_impl.FailedPreconditionError: . is not a directory

weixin_48978546 2021.05.13
if os.path.exists(checkpoint_save_path + '.index'): 您好，麻烦问一，下其中的'.index'作用是什么
- plus_left回复weixin_48978546 2021.05.13
  生成的checkpoint文件里会有这个索引文件，那么存在则证明权重保存成功就是一个看模型参数是否保存的标准