wandb.config

View source on GitHub

Config object

config() -> None

Config objects are intended to hold all of the hyperparameters associated with a wandb run and are saved with the run object when wandb.init is called.

We recommend setting wandb.config once at the top of your training experiment or setting the config as a parameter to init, ie. wandb.init(config=my_config_dict)

You can create a file called config-defaults.yaml, and it will automatically be loaded into wandb.config. See https://docs.wandb.com/library/config#file-based-configs.

You can also load a config YAML file with your custom name and pass the filename into wandb.init(config="special_config.yaml"). See https://docs.wandb.com/library/config#file-based-configs.

Examples:

Basic usage

wandb.config.epochs = 4
wandb.init()
for x in range(wandb.config.epochs):
    # train

Using wandb.init to set config

wandb.init(config={"epochs": 4, "batch_size": 32})
for x in range(wandb.config.epochs):
    # train

Nested configs

wandb.config['train']['epochs] = 4
wandb.init()
for x in range(wandb.config['train']['epochs']):
    # train

Using absl flags

flags.DEFINE_string(‘model’, None, ‘model to run’) # name, default, help
wandb.config.update(flags.FLAGS) # adds all absl flags to config

Argparse flags

wandb.init()
wandb.config.epochs = 4

parser = argparse.ArgumentParser()
parser.add_argument('-b', '--batch-size', type=int, default=8, metavar='N',
                    help='input batch size for training (default: 8)')
args = parser.parse_args()
wandb.config.update(args)

Using TensorFlow flags (deprecated in tensorflow v2)

flags = tf.app.flags
flags.DEFINE_string('data_dir', '/tmp/data')
flags.DEFINE_integer('batch_size', 128, 'Batch size.')
wandb.config.update(flags.FLAGS)  # adds all of the tensorflow flags to config