Skip to content

Experiment Tracking

Configure loggers to track experiments.

TensorBoard (Built-in)

trainer:
  logger:
    _target_: pytorch_lightning.loggers.TensorBoardLogger
    save_dir: logs
    name: my_experiment
    version: null  # Auto-incrementing version

# View logs
tensorboard --logdir logs

Weights & Biases

trainer:
  logger:
    _target_: pytorch_lightning.loggers.WandbLogger
    project: my_project
    name: experiment_name
    save_dir: logs

MLflow

trainer:
  logger:
    _target_: pytorch_lightning.loggers.MLFlowLogger
    experiment_name: my_experiment
    tracking_uri: file:./mlruns

CSV Logger

trainer:
  logger:
    _target_: pytorch_lightning.loggers.CSVLogger
    save_dir: logs
    name: my_experiment

Multiple Loggers

trainer:
  logger:
    - _target_: pytorch_lightning.loggers.TensorBoardLogger
      save_dir: logs
    - _target_: pytorch_lightning.loggers.WandbLogger
      project: my_project

For advanced logging features, see PyTorch Lightning Logger docs.