Ask HN: Distributed M1/2 Training

5 points by aethelyon ↗ HN
Are there any good benchmarks or examples using Apple M1 or M2 in distributed training or inference? With the new M2 Mac Pro hardware it seems like there’s an alternative to H100s.

2 comments

[ 4.4 ms ] story [ 16.0 ms ] thread
> With the new M2 Mac Pro hardware

Even just the Studios, training could be really compelling with the shared memory.

seems like it, but no one is talking about it – everyone I ask IRL says performance is bad, but not seeing in benchmarks