Rename `XUMX-L` to `XUMX-M`
As UMX-L and XUMX-L are actually trained on different-sized datasets, we rename XUMX-L to XUMX-M to indicate that it was trained on a smaller dataset.
As UMX-L and XUMX-L are actually trained on different-sized datasets, we rename XUMX-L to XUMX-M to indicate that it was trained on a smaller dataset.