Weights & Biases’s documentation

This package provides an interactive CLI and python client for wandb.ai.

Quickstart

If you want to work with private buckets or push data to the cloud you must first signup

$ pip install wandb
$ cd my_training_dir
$ wandb init
$ ./my_training.py arg1 | wandb my_bucket model.json weights.h5

To manually push files to the cloud run:

$ wandb push my-new-bucket somefile.proto

To pull down a public model or sync your local directory with the cloud you can run:

$ wandb pull zoo/inception_v4

You can access configuration and push or pull via the API directly in your python scripts:

import wandb
conf = wandb.Config()
client = wandb.Api()
client.pull("zoo/inception")