Comparison with other tensorboard libs

[DrJimFan]

Hi, sorry for the non-coding question. What's the difference between this library and the other older ones like Crayon and tensorboard_logger? Crayon looks a bit heavy-weight because it requires docker. tensorboard_logger has not been maintained for 9 months. Other than these, are there any functional differences?

[ethancaballero]

I think tensorboard_logger can only be used for scalar data, so it can't be used for data with greater than 0 dims such as histograms.

This tensorboard-pytorch repo seems to be best because it retains all of tensorboard's functionality while remaining lightweight.

[lanpa]

tensorboard_logger only saves scalar values, while tensorboard-pytorch saves other data such as image, audio, text, embedding, etc. Another difference is the suffix "-pytorch", tensorboard-pytorch let you pass pytorch tensors directly and it will make conversion behind. Compared with crayon, this package also supports more data format.

[DrJimFan]

@lanpa this is awesome. I'll stick with your library for my research. Thanks for the efforts!