Track experiment logs with Cascade#

Minimal setup#

Suppose you already have a python script with some prints or other logging solutions. Here is the minimal script with two lines of log.

if __name__ == "__main__":
    print("Hello!")
    print("I am example log")

Track your logs#

We will modify this script so that at each run it will save a new model to a line of experiments with a log file.

from cascade.lines import ModelLine
from cascade.models import BasicModel

if __name__ == "__main__":
    print("Hello!")
    print("I am example log")

    model = BasicModel()
    model.add_log()

    line = ModelLine("log_tracking_example")
    line.save(model)

Notice that add_log does not accept any arguments. This is because the script now expects to be run using cascade run that will capture the logs and do all the magic necessary.

In the terminal we can do:

cascade run track_logs.py --log

After that the special folder will be created that stores our logs automatically. You can find the log files in <line_folder>/<model_folder>/files/cascade_run.log

.
├── 00000
│   ├── artifacts
│   ├── files
│   │   └── cascade_run.log
│   ├── meta.json
│   ├── model.pkl
│   └── SLUG
└── meta.json