By setting the maxRecordsPerFile option when writing the DataFrame, which controls the maximum number of records per file, thus influencing the size of part-files.

By using coalesce to reduce the number of partitions before writing, which directly controls the number of part-files and their sizes.

By repartitioning the DataFrame based on a specific column, which can help in creating balanced part-files but does not directly control their size.

By increasing the shuffle partitions, which indirectly affects the part-file sizes by distributing data more evenly across partitions.