Fork an experiment

Forking refers to creating a run based on an existing run. This way, you can restart a run at a particular step.

To fork an experiment, create a run and specify the parent run and step to fork from:

run = Run(
    run_id="adventurous-barracuda",
    experiment_name="swim-further",
    fork_run_id="likable-barracuda",
    fork_step=102,
)

Forking and inheritance

The new run inherits all metrics up to and including the fork point specified by the fork_step argument.

A forked run inherits the following metadata:

Metadata	Attribute type or path	How you log it
Metrics	FloatSeries attributes	run.log_metrics=(...)
Tags	StringSet	run.add_tags(tags=...) or via tags management in the app
Description	sys/description	run.log_configs(data={"sys/description": "..."}) or in the app

Setting step for metrics

When specifying a step for the metrics of the fork run, the value must be higher than the fork step.

run = Run(
    ...,
    fork_step: 100,
)

for step in epoch:
    # your training loop
    run.log_metrics(
        data={"loss": 0.11, "acc": 0.81},
        step=101,                              # or 100.1, for example
    )