You have two different voltages to deal with. Some approaches follow. I'm writing more about one of them because I bothered to provide a circuit. My apologies in that regard. It should not be taken to mean I prefer it. It just needs more of an explanation. The other two options are fine.
- Two supplies: a \$3.3V\cdot 3=9.9V\$ supply, plus some extra voltage headroom plus two separate load resistors to help balance out the currents; and a \$3.3V\cdot 4=13.2V\$ supply, again plus some extra voltage headroom plus two separate load resistors to help balance the currents. You might use one primary supply, plus two of those cheap buck converters and adjust them to provide a reasonable rail and then drop some voltage from there, with resistors. You can tweak them for the two different rails easily.
- Single supply: Enough for at least \$13.2V\$ plus some extra voltage headroom, plus four separate load resistors to help balance the currents out. This would dissipate more wasted power in the load resistors than solution (1) above. Here you could just use the primary supply.
- Place all four modules in series with each other, since they all require the same current. This will require \$(8+6)\cdot 3.3V=46.2V\$, plus some voltage headroom for current regulation. Since \$48VDC\$ @ \$1A\$ supplies are often available for around US$20 on ebay, this might be an approach. You'd need a regulator, which could be composed of something like this:

simulate this circuit – Schematic created using CircuitLab
You can remove and therefore bypass \$R_3\$, but you'll need to dissipate up to \$7W\$ in \$Q_1\$, then. Your call where to put the power. Note that we are talking about tossing away about \$8W\$ in the control circuit. But your load is about \$14W\$, so perhaps it is tolerable.
There's not a lot of headroom here. So if the required voltage varies too much, it may not perform as expected with these values as shown. There is about \$1 \textrm{V}\$ of headroom in \$R_1\$, so that's all there is to play with for the LEDs. But the circuit can be better adjusted, if you know exactly what the series requirement is.
EDIT: Tony has brought up the idea that \$R_3\$ isn't necessarily a great choice. He's right. But \$R_3\$ is completely unnecessary. It's there for one purpose and one purpose only -- to share some of the dissipation. I sized it as large as it should be, so drop it down from there as needed. Down to \$0\Omega\$ is fine. It can be completely removed, in fact. But the BJT will need to dissipate whatever \$R_3\$ doesn't. Just a way to trade off where power is dissipated, is all. It serves no other important purpose here.