Using my dataset works poorly

[KiligXiao]

老师您好，我是一位新手，我在使用我自己的数据时，效果很不理想，这是我自己创建的数据集
[{'target': array([[ -1.7835388, -1.6203486, -1.5254043, ..., -2.3906047, -2.3321326, -2.1873922], [ -2.420903 , -2.3120046, -2.2685359, ..., -2.660461 , -2.6904154, -2.6093047], [ -3.283385 , -3.2697637, -3.1871657, ..., -3.82963 , -3.888748 , -3.9434862], ..., [-17.851435 , -17.806448 , -17.971878 , ..., -18.524988 , -18.615858 , -18.53719 ], [-18.293644 , -18.29638 , -18.415382 , ..., -18.923132 , -18.889847 , -18.922173 ], [-19.470575 , -19.23762 , -19.259882 , ..., -20.00108 , -20.227856 , -20.32058 ]], dtype=float32), 'start': Period('2024-02-17 13:00:00', 's')}]
并在训练时把input_size设置为了上述中dataset_train[0]["target"]的行数，虽然数据是只有一维，但是我把时间序列分为了23组（23个小时）
estimator_rc = TimeGradEstimator(
input_size=23,
hidden_size=64,
num_layers=2,
dropout_rate=0.1,
lags_seq=[1],
scheduler=scheduler,
num_inference_steps=149,
prediction_length=24,
context_length=24,
freq="1s",
scaling="std",
trainer_kwargs=dict(max_epochs=5, accelerator="gpu", devices="1"),
)请问哪里需要修改的呢