"First-class and native support for visualization and model debugging with TensorBoard"
This is huge. TensorBoard makes it so much easier to understand what's going on when you're training a model over hours or days, and need to tweak hyperparameters.
This is huge. TensorBoard makes it so much easier to understand what's going on when you're training a model over hours or days, and need to tweak hyperparameters.
I was recently applying NVidia's stylegan model (https://github.com/NVlabs/stylegan) to a different type of image. Their model uses TF and supports TensorBoard out of the box, so I didn't even try to use pytorch, even though others have shared their implementations (e.g. https://github.com/rosinality/style-based-gan-pytorch).