Does DeepAR need features and target to be normalized

I am predicting sales (say dollars, which is the target) using DeepAR by training multiple related time series. Following are details of features and target (value ranges):

Target variable: (sales in dollars) typical values range from 3000 to 40000 (avg: 22000)

Dynamic Feature 1: values range from 1 to 50 (int)
Dynamic Feature 2: values range from 1 to 10 (int)
Dynamic Feature 3: possible values are 0 and 1 (int/binary)
Dynamic Feature 4: values range from 1 to 5 (int)
Dynamic Feature 5: values range from 0 to 1 (float)

following are cat features' values after encoding them to integer from string values,

Categorical Feature 1: values range from 0 to 2
Categorical Feature 1: values range from 0 to 17
Categorical Feature 1: values range from 0 to 153
Categorical Feature 1: values range from 0 to 1378
Categorical Feature 1: values range from 0 to 44
Categorical Feature 1: values range from 0 to 2

Considering above value ranges for my features and target, is it recommended to use any kind of data normalization/scaling or this is something that DeepAR manages on its own. If yes, which normalization technique works best with DeepAR and will it be required for both features and target (to bring them both on same scale) or only on the features. Thanks in advance.

No file, test, or notebook entry point is identified in the issue. Start with the DeepAR documentation and the SageMaker example materials, then determine whether normalization guidance is documented for the target, dynamic features, and categorical features. Done means providing a clear, referenced recommendation for the reported value ranges.

Written by the indexing model from the issue text.