[ENH] Standardize TimeXer v1 output shape to (batch_size, prediction_length, 1) for point predictions #1932
Description
Is your feature request related to a problem? Please describe.
Currently, TimeXer handles point predictions by outputting a 2d tensor (batch_size, prediction_length) for a single target and a list of this 2d tensor in case of multiple targets. The scope of standardizing this model's output has been determined as a necessity.
Describe the solution you'd like
A simple fix which adds an extra dimension to the existing 2d tensor would suffice.