WandbCallback

class trove.evaluation.callback_wandb.WandbCallback

__init__()

Simple callback to log metrics to Wandb.

setup(run_name=None, args=None, model=None, init_kwargs=None, extra_configs=None, **kwargs)

init a wandb run.

Parameters:

• run_name (Optional[str]) – human readable run name.

• args (Optional[Any]) – instance of RetrievalEvaluator.args

• model (Optional[Any]) – model that is being evaluated

• init_kwargs (Optional[Dict]) – additional kwargs for wandb.init() method

• extra_configs (Optional[Union[Dict, List[Dict]]]) – additional configs to be merged with experiment config when initializing wandb.

• **kwargs – Not used. Just to swallow extra arguments.

Return type:

None

log(*args, **kwargs)

Log results to wandb run.