Error : Train TimeGradEstimator with ComplexSeasonalTimeSeries from GluonTS

[c1adrien]

Hello,

Thank you for your well-written code. I want to train TimeGrad with the data generated by GluonTS ComplexTimeSeries.

from gluonts.dataset.artificial import ComplexSeasonalTimeSeries
from gluonts.dataset.common import ListDataset

dataset  = ComplexSeasonalTimeSeries(
    num_series=10,
    prediction_length=21,
    freq_str="H",
    length_low=30,
    length_high=200,
    min_val=-10000,
    max_val=10000,
    is_integer=False,
    proportion_missing_values=0,
    is_noise=True,
    is_scale=True,
    percentage_unique_timestamps=1,
    is_out_of_bounds_date=True,
)

train_ds = ListDataset(dataset.train, freq=dataset.metadata.freq)

The metadata :

dataset.metadata

Output : MetaData(freq='H', target=None, feat_static_cat=[], feat_static_real=[], feat_dynamic_real=[], feat_dynamic_cat=[], prediction_length=21)

and I got

estimator = TimeGradEstimator(
    target_dim=1,
    prediction_length=dataset.metadata.prediction_length,
    context_length=dataset.metadata.prediction_length,
    cell_type='GRU',
    input_size=1484,
    freq=dataset.metadata.freq,
    loss_type='l2',
    scaling=True,
    diff_steps=100,
    beta_end=0.1,
    beta_schedule="linear",
    trainer=Trainer(device=device,
                    epochs=3,
                    learning_rate=1e-3,
                    num_batches_per_epoch=100,
                    batch_size=64,),
    max_idle_transforms=100,
)
predictor = estimator.train(dataset.train, num_workers=8)

Output : gluonts.exceptions.GluonTSDataError: Input for field "target" does not have the requireddimension (field: target, ndim observed: 1, expected ndim: 2)

I am not exactly sure how to fix the problem and I am having trouble understanding what target_dim expects. Indeed, this target_dim is not present in other models such as DeepAREstimator. Thank you in advance for your help!