Running data loader workers on GPU

deployment
1

Hi all !

I use below data loaders. How can ensure that workers of the data loaders run on GPU ?

I’m on below package versions.

torch 2.0.1
pytorch-forecasting 1.0.0

# Training data set timeseries
training_data_timeseries = TimeSeriesDataSet(
    train_data,
    time_idx="Timestamp",
    target="Target Variable",
    group_ids=["Timeseries"],
    min_encoder_length=max_encoder_length // 2,
    max_encoder_length=max_encoder_length,
    min_prediction_length=max_prediction_length // 2,
    max_prediction_length=max_prediction_length,
    time_varying_unknown_reals=['Target Variable'],
    add_relative_time_idx=True,
    add_target_scales=True,
    add_encoder_length=True,
    target_normalizer=None
)

# Validation data set timeseries
validation_data_timeseries = TimeSeriesDataSet.from_dataset(training_data_timeseries, train_data, predict=True, stop_randomization=True)


# Create dataloaders for our model
train_dataloader = training_data_timeseries.to_dataloader(train=True, batch_size=batch_size, num_workers=workers, pin_memory=True)
val_dataloader = validation_data_timeseries.to_dataloader(train=False, batch_size=batch_size, num_workers=workers, pin_memory=True)

Thanks
Priyanka