A tool which helps you to release the true potential of Tensorboard.
You can keep almost everything under a single Tensorboard, which can be really useful for presentations, tutorials, experiment results.
- Text logger
- Continuous text logger
- Scalar logger
- Histogram logger
- Image logger
pip install git+https://github.com/gaborvecsei/Swiss-Army-Tensorboard.git
(Or you can clone the repo and python setup.py install)
There is an example for every logger inside logger_examples.py.
Just a quick snippet how easy to use this package:
import numpy as np
from swiss_army_tensorboard import tfboard_loggers
scalar_logger = tfboard_loggers.TFBoardScalarLogger("./log_folder")
for i, t in enumerate(np.arange(0.0, 1.0, 0.01)):
val = np.sin(2 * np.pi * t)
scalar_logger.log_scalar("scalar_tag", val, i)Text logger output (inside Tensorboard):
Image and Histogram loggers are made based on the following gist:
[1] https://gist.github.com/gyglim/1f8dfb1b5c82627ae3efcfbbadb9f514
Keras train-validation callbacks is based on this stackoverflow answer:
[2] https://stackoverflow.com/a/48393723/5108062
Gábor Vecsei