labml 0.4.30

Organize Machine Learning Experiments

LabML

LabML is a library to track PyTorch experiments.

LabML keeps track of every detail of the experiments: source code, configurations, hyper-parameters, checkpoints, Tensorboard logs and other statistics. LabML saves all these automatically in a clean folder structure.

This is an example usage of Tracker

from labml import monit, tracker

for epoch in monit.loop(50):
for i in monit.iterate("Train", 10):
    time.sleep(1e-2)
    loss = 50 - epoch + np.random.randint(100) / 100
    tracker.save('loss.train', loss)

if (epoch + 1) % 5 == 0:
    logger.log()

Here’s the output,

Create an experiment and save the configurations with a couple of lines of codes,

from labml import experiment

experiment.create(name='sin_wave')
experiment.configs(configs)
experiment.start()

View all your experiments locally with Dashboard:

You can also monitor your experiments on Slack. When configured you will be receiving updates like following on a Slack thread. Join our Slack workspace to see samples.

Installation

pip install labml

Links

💬 Slack workspace for discussions

📗 Documentation

📑 Articles & Tutorials

👨‍🏫 Samples

Citing LabML

If you use LabML for academic research, please cite the library using the following BibTeX entry.

@misc{labml,
 author = {Varuna Jayasiri, Nipun Wijerathne},
 title = {LabML: A library to organize machine learning experiments},
 year = {2020},
 url = {https://lab-ml.com/},
}