Edit run configs
To overwrite single values logged with the log_configs()
method, assign a new value to the same attribute path. The new value overwrites the previous one.
For example, to edit a run's learning rate:
run = Run(
run_id="seagull-s68kj78", # a run with this ID already exists
)
# Attribute "parameters/learning_rate": 0.001" was logged previously
run.log_configs({"parameters/learning_rate": 0.002})
run.close()
Updating an experiment with new configs
Instead of modifying individual runs, you can map runs to a single experiment and create a new run each time a configuration changes. To learn more, see Fork an experiment.