Discrepancies between hugging face icon-eu data and data downloaded from consumer

[gabrielelibardi]

If I download icon-eu data with the consumer I get something like this, where some variables are split over the pressure levels

Variables list:
1. 51_t
2. 52_t
3. 53_t
4. 54_t
5. 55_t
6. 56_t

...
but in the data I download from hugging face we keep the isobaricInhPa coordinates

Dimensions:        (step: 93, latitude: 657, longitude: 1377, isobaricInhPa: 4)
Coordinates:
  * isobaricInhPa  (isobaricInhPa) float64 32B 900.0 925.0 950.0 1e+03
  * latitude       (latitude) float64 5kB 29.5 29.56 29.62 ... 70.38 70.44 70.5
  * longitude      (longitude) float64 11kB -23.5 -23.44 -23.38 ... 62.44 62.5
  * step           (step) timedelta64[ns] 744B 00:00:00 ... 5 days 00:00:00
    time           datetime64[ns] 8B ...
    valid_time     (step) datetime64[ns] 744B dask.array<chunksize=(93,), meta=np.ndarray>
Data variables: (12/19)
    alb_rad        (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    asob_s         (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    aswdifd_s      (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    aswdir_s       (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    clch           (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    clcl           (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    ...             ...
    t_2m           (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    tot_prec       (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    u              (step, isobaricInhPa, latitude, longitude) float32 1GB dask.array<chunksize=(37, 4, 326, 350), meta=np.ndarray>
    u_10m          (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
    v              (step, isobaricInhPa, latitude, longitude) float32 1GB dask.array<chunksize=(37, 4, 326, 350), meta=np.ndarray>
    v_10m          (step, latitude, longitude) float32 337MB dask.array<chunksize=(37, 326, 350), meta=np.ndarray>
Attributes:
    Conventions:             CF-1.7
    GRIB_centre:             edzw
    GRIB_centreDescription:  Offenbach
    GRIB_edition:            2
    GRIB_subCentre:          255
    history:                 2024-04-08T09:11 GRIB to CDM+CF via cfgrib-0.9.1...
    institution:             Offenbach

Is there a way to achieve the same with the consumer? @devsjc @peterdudfield