reference/generated/torchrl.trainers.algorithms.configs.instantiate_trainer

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

instantiate_trainer

torchrl.trainers.algorithms.configs.instantiate_trainer(cfg: DictConfig, *, overrides: Sequence[str] | None = None) → Trainer

Instantiate cfg.trainer, resuming from cfg.resume when it is set.

Without resume this is hydra.utils.instantiate() on cfg.trainer plus registration of the composed configuration on the trainer checkpoint. With resume set to a checkpoint or a CheckpointRotation directory, resume_config() rebuilds the configuration from the saved one with the current command-line overrides on top; the saved logger run is reattached before the logger is constructed (W&B resumes the saved id with resume="must", CSV and TensorBoard keep the saved directory); checkpoints keep accumulating in the resumed rotation directory unless checkpoint_rotation.directory is overridden; and load_from_file() restores the trainer state.

Parameters:

• cfg (DictConfig) – the composed Hydra configuration. It must hold a trainer node and may hold resume.

• overrides (Sequence[str], optional) – command-line overrides applied over the saved configuration. Defaults to the task overrides of the current Hydra run, or none outside a Hydra application.

Returns:

The instantiated trainer, restored from the checkpoint when resuming.

Examples

>>> @hydra.main(config_path="config", config_name="config", version_base="1.3")  
... def main(cfg):
...     trainer = instantiate_trainer(cfg)
...     with trainer.stop_on_signal():
...         trainer.train()