Succinctly calculate differences between all datetime columns in a row

I have implemented the code to identify the Datetime columns and convert them into DateTime.

def func(df):
    for col in df.columns:
        if df[col].dtype == 'object':
            try:
                df[col] = pd.to_datetime(df[col])
            except ValueError:
                pass
        if df[col].dtype == 'datetime64[ns]':
            df = df.diff(axis=1)
            return df

I want to make new columns using these date columns which are the difference between these columns taking 2 at a time. (for instance if there is date1, date2, date3 columns, output should be like date1-date2, date2-date3, date1-date3). How can it be implemented? I have tried using diff() but I need a concise answer?