Skip to content

logging

Jonas Schult edited this page Oct 5, 2018 · 1 revision

Logging

Log messages are generated following python's standard logging module. Log messages are printed to the console as well as saved to the disk in the corresponding logging directory. This makes it easy to keep track of the training/testing progress.

Logging Directory

In the logging directory, log messages, the configuration file and the model checkpoints will be saved during training. In case of testing, the results and pointclouds in obj file format will be saved. The directory is time encoded (%Y-%m-%d@%H:%M:%S) followed by a random string and is a subfolder of logs in the root directory.

Best Practice

It is highly encouraged to commit one's code before starting the training. The current git revision hash will be stored in the log-txt file providing very useful information in order to guarantee reproducibility of the results.

Clone this wiki locally