I take the example from https://pytorch.org/tutorials/intermediate/tensorboard_tutorial.html and copy it. I run it but tensorboard does not show any graph (see attached image). I also run it with my own code and get exactly the same.
To show it I use:
tensorboard --logdir=data/ --host localhost --port 8088
Because simply using tensorboard --logdir does not work.
Using writer.add_image() works though. So, could you shed me some light on how to get the graph?
We have similar issues here and here
CC @lanpa, @orionr who might know, what’s going on.
Hi, which version of pytorch are you using? Have you run step 3?
Any errors or warnings?
Based on your description, you might have changed the logging dir. Also make sure that you are passing the correct parameters.
I am using pytorch 1.2. Yes, that was a mistake in my description but the logging dir is the same. In fact, you can know it’s the same because if it weren’t, then I wouldn’t get anything but I still get the images from writer.add_image()
same issue to me. I run the demo from official tutorial VISUALIZING MODELS, DATA, AND TRAINING WITH TENSORBOARD, nothing changed, no graph emerged.
(anaconda 3 environment: python3.7 pytorch 1.2.0 torchvision 0.4.0 tensorboard 1.14.0 )
Same issue here with python3.7.4, pytorch 1.2.0, torchvision 0.4.0, tensorboard 1.14.0
Could you try the nightly build as described in this issue?
@ptrblck Updating Pytorch to nightly build resolved the issue. Thank you very much for the suggestion.