We will support it after an upstream bug being fixed. https://github.com/pytorch/pytorch/pull/20116
@hellock In the mean time, is there any quick solution to use the torch.utils.tensorboard officially supported in PyTorch 1.1.0? Actually I just need a way to access the loss during training, then I would be able to code the rest myself.
@yxchng If you just want to log the loss, you can simply uncomment this line.
I want to visualize the model using TensorboardX too, but encountered with some problems, when will you support it?
Hope this feature added soon, thank you.
Most helpful comment
@yxchng If you just want to log the loss, you can simply uncomment this line.