ltx2.3 peak memory usage

[bil-ash]

Has anyone tried ltx2.3 inference on CPU? If yes, what was the peak memory usage [only CPU, no GPU] and how much time did it take? Also, what CPU and what mode did you use- t2v, i2v or flf2v?

[h3ndrik]

Well, using the Q3_K_S variant of LTX2.3 22b along with Gemma3 12b Q3_K_S makes it fit into the 32GB my laptop has. Q4_K_S results in an OOM kill. It's took 6h on my laptop until it began generating frames and it says 7h per iteration. So I aborted after 24h.