// RUN โ€” FITS

Can the AMD Instinct MI325X run DDPM-IP (CelebA)?

Yes โ€” here's how fast and up to what context, with real measured numbers.

DDPM-IP (CelebA) on the AMD Instinct MI325X

QuantFits?Max contextSpeedBeyond context
FP16 yes 3562k 6406.8-7830.5 t/s ~42.7-52.2 (slow)
Q8 yes 3566k 12813.6-15661 t/s ~85.4-104.4 (slow)
Q4 yes 3568k 22881.4-27966.1 t/s ~152.5-186.4 (slow)

Estimate (memory-bound). Beyond "max context" the KV cache spills to system RAM offload (depends on your system RAM) and speed drops to the "beyond" figure.

FAQ

Can the AMD Instinct MI325X run DDPM-IP (CelebA)?โ–ถ

Yes. At Q4 it generates ~22881.4-27966.1 tok/s and fits up to 3568k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~152.5-186.4 tok/s.

How much context fits?โ–ถ

At Q4, up to about 3568k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.