← Back to API reference

neuralbench.utils.TrainerConfig

pydantic model neuralbench.utils.TrainerConfig

Joint configuration for Trainer and some callbacks.

Fields:

• accelerator (str)

• accumulate_grad_batches (int)

• devices (int)

• enable_progress_bar (bool)

• fast_dev_run (bool)

• gradient_clip_val (float)

• limit_train_batches (int | None)

• limit_val_batches (int | None)

• log_every_n_steps (int)

• mode (str)

• monitor (str)

• n_epochs (int)

• num_nodes (int)

• num_sanity_val_steps (int)

• patience (int)

• precision (str)

• strategy (str)

field n_epochs: int = 100

field enable_progress_bar: bool = True

field log_every_n_steps: int = 20

field fast_dev_run: bool = False

field gradient_clip_val: float = 0.0

field limit_train_batches: int | None = None

field limit_val_batches: int | None = None

field num_sanity_val_steps: int = 2

field accumulate_grad_batches: int = 1

field strategy: str = 'auto'

field precision: str = '32-true'

field accelerator: str = 'auto'

field devices: int = 1

field num_nodes: int = 1

field patience: int = 5

field monitor: str = 'val/loss'

field mode: str = 'min'

build(logger, callbacks, accelerator: str | None = None, devices: int | None = None, num_nodes: int | None = None) → Trainer

← Back to API reference