oMLX has a fun new feature.
You can connect two Macs to run large models that wouldn't normally be executable, or
you can aim for speed improvements.
I was successful in pairing two Macs, an M2 24GB and an M2 16GB.
(It took a while because it didn't work well)
But the model execution doesn't work.
The cluster page keeps reloading, which is annoying...
Since it's still an experimental feature, I'll have to try using it again when it matures a bit more.
Gemini says: Trying to connect M2 + M5 Pro results in performance degradation due to the M2.
0.6.1 throws an error when turboquant is activated.
The PR for the fix has been merged, so I'll have to wait for the next version.
