Migrate from Weights & Biases
To migrate the run metadata from Weights & Biases to Neptune, use the wandb_to_neptune.py
utility script provided in the neptune-ai/scale-examples repository.
The README
file contains prerequisites and complete instructions.
Code mapping
Learn how the Weights & Biases methods map to the Neptune methods.
wandb.init
run = neptune.Run()
See also Create an experiment.
wandb.config.update
run.log_configs()
See also Log configs.
wandb.log
In Neptune, a logging method depends on the metadata type. For example:
- for metrics, use
run.log_metrics()
- for single files, use
run.assign_files()
- for series of files, use
run.log_files()
For a full list, see Attribute types reference.
wandb.login
Logging in through API isn't required. Instead, save your credentials as environment variables.
wandb.finish
run.close()
See also close()
.
scan_history
To query the logged metadata, use the Neptune Fetcher client library.
import neptune_fetcher.alpha as npt
# Fetch experiments table with regex searching
npt.fetch_experiments_table()
# Fetch metric values with regex searching
npt.fetch_metrics()
See also Query metadata.