Camera ready machine learning plots with a Tensorboard-like API
Project description
paperplots
Camera ready machine learning plots with a Tensorboard-like API
Ever felt like the tensorboard summary writer can do a bit more than just letting you monitor your training logs? Wouldn't it be nice if you could also directly get pretty paper-ready versions of your tensorboard logs? Paperplots does just this!
- Record scalars in the exact same way as you do with tensorboard summarywriter
- Save your training data in easy to access pickle files
- Plot runs, whole experiments, and even compare different experiments
- Automatically calculate simple convenience transforms such as rolling averages and shaded error regions
Examples
Writing Logs
writer = Writer(logdir=f'runs/algo', run_name=f'run1')
for i in range(iter):
writer.add_scalar("loss", loss_value, i, ylabel="loss", xlabel="num iter", name="algo loss")
writer.close()
Plotting Written Logs
plotter = Plotter(logdir='runs')
# Plotting runs
plotter.plot_run(run_name="algo/run1")
# Compare two runs
plotter.plot_run(run_name=["algo/run0", "algo/run1"])
# Plot an average over all runs in the experiment
plotter.plot_experiment(exp_name="algo")
# Compare two experiments
plotter.plot_experiment(exp_name=["algo1", "algo2"])
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
paperplots-0.2.0.tar.gz
(5.8 kB
view hashes)
Built Distribution
Close
Hashes for paperplots-0.2.0-py2.py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 9de9764753ff97457be7384a125b43d61b80a4cf3fa4467c6df8b41f0d7e1ddb |
|
MD5 | 679153907a3e8b87b5fa5430359c61f7 |
|
BLAKE2b-256 | e1e9241eae23e18a4de8fef7e778429c79f89e399c9d8cc2c48aa91e875b22fc |