trainerERA5_multistep_sum

trainerERA5_multistep_sum#

Attributes#

logger

Classes#

Trainer

Trainer class for handling the training, validation, and checkpointing of models.

Module Contents#

trainerERA5_multistep_sum.logger#
class trainerERA5_multistep_sum.Trainer(model: torch.nn.Module, rank: int)#

Bases: credit.trainers.base_trainer.BaseTrainer

Trainer class for handling the training, validation, and checkpointing of models.

This class is responsible for executing the training loop, validating the model on a separate dataset, and managing checkpoints during training. It supports both single-GPU and distributed (FSDP, DDP) training.

model#

The model to be trained.

Type:

torch.nn.Module

rank#

The rank of the process in distributed training.

Type:

int

module#

If True, use model with module parallelism (default: False).

Type:

bool

train_one_epoch(epoch, conf, trainloader, optimizer, criterion, scaler, scheduler, metrics)#

Trains the model for one epoch.

Parameters:

  • epoch (int) – Current epoch number.

  • conf (dict) – Configuration dictionary containing training settings.

  • trainloader (DataLoader) – DataLoader for the training dataset.

  • optimizer (torch.optim.Optimizer) – Optimizer used for training.

  • criterion (callable) – Loss function used for training.

  • scaler (torch.cuda.amp.GradScaler) – Gradient scaler for mixed precision training.

  • scheduler (torch.optim.lr_scheduler._LRScheduler) – Learning rate scheduler.

  • metrics (callable) – Function to compute metrics for evaluation.

Returns:

Dictionary containing training metrics and loss for the epoch.

Return type:

dict

validate(epoch, conf, valid_loader, criterion, metrics)#

Validates the model on the validation dataset.

Parameters:

  • epoch (int) – Current epoch number.

  • conf (dict) – Configuration dictionary containing validation settings.

  • valid_loader (DataLoader) – DataLoader for the validation dataset.

  • criterion (callable) – Loss function used for validation.

  • metrics (callable) – Function to compute metrics for evaluation.

Returns:

Dictionary containing validation metrics and loss for the epoch.

Return type:

dict