InvalidCastException (Int32 to Single) when retraining from database

I have my data stored in a single table that I query to grab certain columns (and "de-NULL"). I was previously saving as a CSV (due to another issue re: certificate trust) and training largely successfully. However, as I move into retraining, I'm trying to pull directly from the database with a train/test split. When I do this, I get an InvalidCastException: "Unable to cast object of type 'System.Int32' to type 'System.Single'". Presumably this is due to any number of INT fields that I'm returning (I've got some MONEY, some INT, some DATE, and some NVARCHAR); I think the CSV route is treating everything as a string and parsing to whatever, whereas the DB is typing everything.
My question is - is it up to me to prepare the data with the right types here (either by transform or by having my SQL convert everything to MONEY/NVARCHAR), or is there something I'm missing to get it to handle the conversion automatically?

El issue no menciona archivos ni tests. Empieza reproduciendo la división train/test respaldada por la base de datos con los tipos de columna SQL indicados, y después rastrea dónde se convierte el valor Int32 a Single. Se considera terminado cuando se haya corregido el comportamiento de conversión y se haya verificado con tipos de base de datos mixtos, o cuando se haya establecido claramente la preparación de entrada requerida.

Escrito por el modelo de indexación a partir del texto del issue.