In the first round of hashing for LSR label IP hashing, the algorithm parses down the label stack as described in LSR Label-only Hashing.
When the algorithm reaches the bottom of the stack, it checks the next nibble. If the nibble value is 4, the packet is assumed to be an IPv4 packet and the result of the label hash is fed into another hash along with the source and destination address fields in the IP packet header. If the nibble value is not 4, the algorithm will just use the label stack hash already calculated for the ECMP path selection.
The second round of hashing for LAG reuses the net result of the first round of hashing.