// RUN โ€” FITS

Can the NVIDIA DGX Spark run AWD-LSTM-MoS + dynamic evaluation (PTB, 2018)?

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

AWD-LSTM-MoS + dynamic evaluation (PTB, 2018) on the NVIDIA DGX Spark

QuantFits?Max contextSpeedBeyond context
FP16 yes 1734k 4151.9-4948.1 t/s ~152.1-181.3 (slow)
Q8 yes 1735k 8303.8-9896.3 t/s ~304.2-362.5 (slow)
Q4 yes 1735k 14828.1-17671.9 t/s ~543.2-647.3 (slow)

Estimate (memory-bound). Beyond "max context" the KV cache spills to SSD swap and speed drops to the "beyond" figure.

FAQ

Can the NVIDIA DGX Spark run AWD-LSTM-MoS + dynamic evaluation (PTB, 2018)?โ–ถ

Yes. At Q4 it generates ~14828.1-17671.9 tok/s and fits up to 1735k context; beyond that it spills to SSD swap and slows to ~543.2-647.3 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1735k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.