When using AzureML, it is not clear to what extent Parquet data can be loaded as a tabular dataset – in terms of supported data types. For example, columns with array data (e.g. an array of floats) does not seem to be supported.