Tensorboard checkpoint : Access is denied. ; Input/output error

问题

I am trying to create a tensor board in Jupyter anaconda the following way. The error occurs when write_images = True, otherwise, this code works fine. Any reason why this happens?

log_dir="logs\\fit\\" + datetime.datetime.now().strftime("%Y%m%d-%H%M%S")

tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=log_dir, 
                                                      histogram_freq=1, 
                                                      write_graph = True,
                                                      write_images = False
                                                      update_freq = 'epoch'
                                                      profile_batch = 3,
                                                      embeddings_freq=1
                                                     )

And I get

UnknownError: Failed to rename: logs\20200219-202538\train\checkpoint.tmp67d5ca45d1404cc584a86cf42d2761d3 to: logs\20200219-202538\train\checkpoint : Access is denied.
; Input/output error

Seems to be random on which epoch it occurs.

回答1:

I had something similar, it seems like the path where you want to safe the checkpoint, which is referred to tensorbaord is not available or denied. Do you know colab ? I would suggest you copy your code and run your training up there (only if your dataset isnt too large). You can copy your dataset in your Google Drive and access with colab. If it is working in colab, than you may dont have a problem with your code, but propably with your anaconda restrictions. Mount Google Drive (Colab), Colab basics

I know i couldnt solve your problem, but propably this can help you and boost your learning speed with a juicy free Cloud GPU.

来源：https://stackoverflow.com/questions/60309013/tensorboard-checkpoint-access-is-denied-input-output-error