Could you include the code snippet for reproduction? If it's related to the tutorials, it's also better to open an issue in pytorch/tutorials.

Do you want me to rewrite on pytorch/tutorials? I did.
I think it is about tensorflow(2.0) api changes.

I use the following environment and not problem occurs. The code I used is #30525, which is from the doc as well.
@develooper1994 What is your value of save_path?

torch==1.3.1
torchvision==0.4.2
tensorboard==2.0.2
tensorflow==2.0.0
tensorflow-estimator==2.0.1
python==3.6.9

Still can't produce with 1.4.0.dev20191211 and python 3.7.5

I didn't try after fail. I have started long run.

#27716
This pr seems to be the problem. At first glance I couldn't find any version of tensorflow where tf.io.gfile has these functions... but I assume there is some version of tensorflow that works.

Has anyone got the solution yet I still have the problem

torch 1.5.0.dev2020011
tensorboard 2.0.0
tensorflow 2.0.0
python 3.7

You can create a new environment, then install pytorch and tensorboard. Note: you should not install tensorflow as there would be a kind of conflict with tensorboard.

Thank you @Jean-KOUAGOU this fixed my problem.

As you can see from these two links

https://github.com/tensorflow/tensorboard/blob/master/tensorboard/compat/__init__.py#L46-L57
https://github.com/tensorflow/tensorboard/blob/a72870cb1e59f71f49f72bfd4c2f3ec25dc75fa9/tensorboard/compat/tensorflow_stub/io/gfile.py#L72

tensorboard has it's own implementation of gfile that it uses if tensorflow is not present on the system. @sanekmelnikov used a non-public API from this stub implementation when implementing #27716

Temporary solution

import tensorflow as tf
import tensorboard as tb
tf.io.gfile = tb.compat.tensorflow_stub.io.gfile

I would like to add to @korakot s workaround that it did slow down program execution by a huge factor if you have many summarywriter calls. Has anybody else observed this?

Using latest pytorch 1.4 and tensorboard 2.2

Removing Tensorflow from the Pytorch/Tensorboard env, as mentioned by @Jean-KOUAGOU, worked for me.

@clefourrier , could you clarify what you did? I am having the same problem, thanks.

@korakot , your solution works for me but I don't understand what it specifically does. Could you please explain? Thanks!

@fabrizioschiano : If you created a virtualenv to use pytorch in (using this for example), you have to make sure that you only installed pytorch and tensorboard, but not tensorflow, as it has conflicts with the others. As I said, removing tensorflow from the vritualenv containing pytorch and tensorboard solved the problem for me.

If you did not create a virtualenv and installed everything in the same spot, either

• you do not need tensorflow and should remove it
• you need tensorflow as well as pytorch and should actually use virtual environments to manage your python dependencies (good practice anyway)

Had the same issue with google colab. Uninstalling tensorflow will remove the %tensorboard magic, so you need to keep it if you're using tensorboard inline in a jupyter cell . @korat's solution worked