windows环境下使用python中tensorflow的tensorboard功能无法创建指定路径的问题
An error occurred while assigning a directory path to tensorboard:
tensorflow.python.framework.errors_impl.InvalidArgumentError: Failed to create a directory: F:\tensorflow\catdog\logs/CatsDogsCNN64*2-1602584133; Invalid argument
code:
1 NAME = f'CatsDogsCNN64*2-{int(time.time())}' 2 logD = "F:\\tensorflow\\catdog\\logs\\{}".format(NAME) 3 tsrBd = TensorBoard(log_dir=logD) 4 print(logD)
model.fit(X, y, batch_size=32, epochs=10, validation_split=0.1, callbacks=[tsrBd])
I've tried a bunch of methods to create the path string including the r-literal, the double back slash '\\' and os.path.join().
None of those worked. See the issue: https://github.com/tensorflow/tensorboard/issues/2023
But the solution in this issue does not work for me.
The solution I found is that you should not provide a dynamic path when assigning log_dir to tensorboard but rather use a fixed name (string), such as:
logD = os.path.join("logs", "c&d")
So this is bascially the same as creating manually a directory in windows.
Hopefully google can fix this bug for Windows in the future. It is extremely frustrating and annoying to deal with this kind of problem because you have to pour so much energy in debugging and gain little from it.
