Lorenzo Stella

@strakehyr looking at the `TemporalFusionTransformerEstimator` code, I can see that it requires one to explicitly set the names (and dimensions) of additional feature fields in the data. For example, for...

@strakehyr you may want to try to configure the estimator as in the following example: ```python from typing import List from gluonts.dataset.common import ListDataset from gluonts.mx import TemporalFusionTransformerEstimator dataset =...

> Something I'm wondering: If we merge this PR we break the assumption that time-series are regular. Is that an assumption we rely on in other places? The "uniform sampling"...

> Single letter variable names. While I agree in general, I think extremely local scopes like comprehensions may benefit from short names. This said, I can change them :-)

Make sure to install `black==22.1.0` and to run ``` black src test ``` before committing your code, so that it takes care of some styling issues

There's also a few complaints from `flake8`, see the bottom of [this log file](https://pipelines.actions.githubusercontent.com/LG0kBijBY35Fqw47Be7JAOcSx2cowoyb7WTTLj8BI5oz3i77sX/_apis/pipelines/1/runs/19695/signedlogcontent/2?urlExpires=2022-03-18T11%3A37%3A32.3331116Z&urlSigningMethod=HMACV1&urlSignature=kDh3%2FuBNnmGyS38YJzZCZSU5JQ8%2BYrQSOeS2ivkhFSQ%3D)

@strakehyr could you expand on this idea? Where would you think it is best to specify all the additional features? In separate dataframes?

So you would provide multiple dataframes, specifically: - a “target” one, shape `time_length x num_series` - one for static features, shape `num_series x num_features` (each column with its own `dtype`,...

@mvanness354 thank you for spotting this! The change you propose makes sense to me, assuming that `locationID`s are indeed distinct across the dataset. Do you want to open a PR...

@jaheba as a fix, would it make sense to bake `.to_timestamp` in any utils (hopefully it’s a single place) we use to turn `DataEntry` into pandas?