Updating Meta in "BandPadding()" in data_preprocessor.py - overwrites all values with 0, instead of only padded ones

[AngNB]

The code for BandPadding() currently overwrites all values in the meta with '0'. (Tested by printing min/max values after and before data preprocessing steps; only with "ResizeToEncoder()", "BandFilter()" and "BandPadding()").

Codeblock currently:

def update_meta(self, meta):
    """Tracking the meta statistics/info for next processor."""
    meta["data_bands"] = meta["encoder_bands"]
    for k in self.avail_bands_mask.keys():
        size = self.avail_bands_mask[k].shape
        meta["data_mean"][k] = torch.full(
            size, fill_value=self.fill_value, dtype=torch.float
        )
        meta["data_std"][k] = torch.ones(size, dtype=torch.float)
        meta["data_min"][k] = torch.full(
            size, fill_value=self.fill_value, dtype=torch.float
        )
        meta["data_max"][k] = torch.full(
            size, fill_value=self.fill_value, dtype=torch.float
        )
        if self.used_bands_indices[k] is not None:
            meta["data_mean"][k][self.avail_bands_mask[k]] = meta["data_mean"][k][
                self.used_bands_indices[k]
            ]
            meta["data_std"][k][self.avail_bands_mask[k]] = meta["data_std"][k][
                self.used_bands_indices[k]
            ]
            meta["data_min"][k][self.avail_bands_mask[k]] = meta["data_min"][k][
                self.used_bands_indices[k]
            ]
            meta["data_max"][k][self.avail_bands_mask[k]] = meta["data_max"][k][
                self.used_bands_indices[k]
            ]
    return meta

Changes proposed:
Save data values, before overwriting with '0's