Heterogenous vGPU profiles on Two nodes of a VMware cluster

Hello Team,

Two node VMware Cluster. Each node has single A100 cards. Running MIG on each host. Install ESX 7.0 UPdate 2. Since the process for creating GPU Instances and Compute instances is now automated with vSphere 7 Update 2 and later, so created 2 VM with 3g,20GB profile on one host and 3VM with 2g.10gb on another Host. Live migration doesn’t work . Getting error as "one or more devices (pcipassthru0) required by vm xx-xxx are not available on host xx.xx.xx.xx.

Is this supported to have different vGPU profiles on different nodes in Cluster? is the above design valid
If yes, any specific rule to follow.
Any suggestion for the error message