Create a detached run
To create a run that's not part of an experiment lineage, use the Run
constructor and omit the experiment_name
parameter:
from neptune_scale import Run
if __name__ == "__main__":
run = Run()
The line
if __name__ == "__main__":
ensures safe importing of the main module. For details, see the Python documentation.
To take advantage of forking and analysis of full run history, create an experiment rather than a detached run.
Then, to track metadata, call a logging method on the Run
object:
for step in epoch:
# your training loop
acc = ...
loss = ...
run.log_metrics(
data={"metrics/accuracy": acc, "metrics/loss": loss},
step=1,
)
To get a link to the run in the Neptune web app:
from neptune_scale import Run
run = Run(...)
run.get_run_url()
For details, see Construct Neptune URLs.
Set custom run ID
To override the auto-generated run identifier with a custom one, pass a string to the run_id
argument:
run = Run(run_id="astute-kittiwake-23")
The custom run ID provided to the
run_id
argument must be unique within the project. It can't contain the/
character.
View runs in the app
In the All runs tab, by default, Neptune displays the latest run of each experiment.
To show all runs, switch from Experiments to Runs:
You can save your table configuration as a custom view, which makes it easier to work on the runs later.