Configure Your Lightning Trainer

Under the framework of PyTorch Lightning, we use the Lightning Trainer object for all stuff related to training, such as number of epochs, training strategy, device, etc. Just like optimizer, each task is supposed to be given a trainer for training. We could either use a uniform trainer across all tasks or assign a distinct trainer to each task.

To configure the Lightning trainer for your experiment, link the trainer field in the main YAML file to sub-config file in the sub-directory /trainer/, and specify the class lightning.Trainer and its arguments in the sub-config. Note that there is only one type of Trainer class i.e. lightning.Trainer that we specify. Here is an example:

./clarena/example_configs
├── __init__.py
├── entrance.yaml
├── experiment
│   ├── example.yaml
│   └── ...
├── trainer
│   └── cpu.yaml
...

example_configs/experiment/example.yaml

defaults:
  ...
  - /trainer: cpu.yaml
  ...

example_configs/trainer/cpu.yaml

_target_: lightning.Trainer

default_root_dir: ./outputs/${experiment_name}/${now:%Y-%m-%d_%H-%M-%S}

log_every_n_steps: 500

accelerator: cpu
devices: 1

max_epochs: 9

Supported Training Options

Please refer to Lightning documentation for all supported config fields (arguments) for the trainer.