Migrate from Neptune Legacy (2.x)
This guide shows how to:
- Copy run data from Neptune Legacy (
app.neptune.ai
) to Neptune (scale.neptune.ai
). - Update your code from
neptune
toneptune-scale
.
Copy runs from Neptune Legacy
To copy runs from Neptune 2.x
to Neptune 3.x
, use the run migration utility.
Migrate from Neptune Legacy Python client
Learn about the code changes that you must implement in your existing scripts before migrating from version 1.x
of the Neptune client library to neptune-scale.
Installation
neptune 1.x
pip install neptune
neptune-scale
pip install neptune-scale
Import
neptune 1.x
import neptune
neptune-scale
import neptune_scale
Run operations
For details on all parameters and methods, see Run.
Initialize a run
neptune 1.x
run = neptune.init_run()
neptune-scale
run = neptune_scale.Run()
sync()
neptune 1.x
run.sync()
neptune-scale
run.wait_for_processing()
wait()
neptune 1.x
run.wait()
neptune-scale
run.wait_for_submission()
Stop a run
neptune 1.x
run.stop()
neptune-scale
run.close()
Logging metadata
For details on supported metadata types, see Attribute types reference.
Add tags
neptune 1.x
run["sys/tags"].add(["tag1", "tag2", ...])
neptune-scale
run.add_tags(["tag1", "tag2", ...])
Add group tags
neptune 1.x
run["sys/group_tags"].add(["tag1", "tag2", ...])
neptune-scale
run.add_tags(["tag1", "tag2", ...], group_tags=True)
Log a dictionary of values
neptune 1.x
run["config"] = stringify_unsupported(config_dict)
neptune-scale
run.log_configs(
data={"config": config_dict},
flatten=True, # handles nested structures
cast_unsupported=True, # enables type casting
)
Log metrics
neptune 1.x
run["metrics"].append({"acc": acc, "loss": loss, ...})
neptune-scale
run.log_metrics(
data={"metrics/acc": acc, "metrics/loss": loss, ...},
step=step
)
note
log_metrics()
requires the step
parameter. An option to automatically increase the step
value isn't supported yet.
Log files
Upload a single file:
neptune 1.x
run["sample"].upload("my_file.png")
neptune-scale
run.assign_files("my_file.png")
Upload a series of files associated with steps:
neptune 1.x
run["predictions"].append("my_file.png")
neptune-scale
run.log_files(files={predictions: "my_file.png"}, step=step)
Log series of text
neptune 1.x
run["logs"].append("some_string")
neptune-scale
run.log_string_series("some_string")