WandBLogger

class composer.loggers.WandBLogger(project=None, group=None, name=None, entity=None, tags=None, log_artifacts=False, rank_zero_only=True, init_kwargs=None)

Log to Weights and Biases.

Parameters

• project (str, optional) – WandB project name.

• group (str, optional) – WandB group name.

• name (str, optional) – WandB run name. If not specified, the State.run_name will be used.

• entity (str, optional) – WandB entity name.

• tags (list[str], optional) – WandB tags.

• log_artifacts (bool, optional) – Whether to log artifacts (Default: False).

• rank_zero_only (bool, optional) –

  Whether to log only on the rank-zero process. When logging artifacts, it is highly recommended to log on all ranks. Artifacts from ranks ≥1 will not be stored, which may discard pertinent information. For example, when using Deepspeed ZeRO, it would be impossible to restore from checkpoints without artifacts from all ranks (default: True).

• init_kwargs (dict[str, Any], optional) – Any additional init kwargs wandb.init (see WandB documentation).

can_upload_files()

Whether the logger supports uploading files.