top | item 47205932

(no title)

elcritch | 2 days ago

Running inference requires sharing intermediate matrix results between nodes. Faster networking speeds that up.

discuss

order

wokkel|2 days ago

I read (but cannot find this anymore) that the information sent from layer to layer is minimal. The actual matrix work happens within a layer. They are not doing matrix multiplication over the netwerk (that would be insane latency wise).

elcritch|23 hours ago

The LLM/transformers attention layers require an O(n^2) operation between all tokens, which does require significant bandwidth.

Yes the latency hurts performance, that why it’s only achieving ~8tok/s.