Creating an Input Schema for Multiple DataFrames in MLflow

[ecram]

Hi everyone,

I am working with MLflow version 2.5.0 and need to create an input_schema for my model. My data schema is divided into several DataFrames, for example:

{
"dataframe_split": {
     "columns": ["ClientGuid", "Instance", "TypeScore", ...],
     "data": [["26c5f035-cf57-4cb7-b67c-55da528ce3e0", "1", "POST", ...]]
},
"dataframe_loan_split": {
     "columns": ["System", "Entity", "Credit", ...],
     "data": [["R", "Entity1", "M1 - MIC. IND. DEB. GAR.", ...]]
},
...
}

Could someone provide an example of how to define an input_schema using MLflow for this type of structure? I am looking for a way to specify all the DataFrames within the input schema.

Thanks for your help.

[Kaniz_Fatma]

Hi @ecram, To create an input schema for multiple DataFrames in MLflow, you can use the mlflow.types.schema module to define the schema for each DataFrame and then combine them.