Duplex basecalling running time

[MDW]

I tried to run bonito in duplex mode:

bonito \
   duplex \
   --pairs pseudopairs.txt \
   --max-cpus 1 \
   --device cuda:0 \
   dna_r10.3 \
   fast5_dir \
   > duplex.fasta

with 780 pairs using Bonito v0.4.0. I used the following GPU card: NVIDIA Corporation TU104GL [Tesla T4] (1 GPU, 64 cores, 256 Gb RAM). It did not finish after 3h.

[downloaded dna_r10.3.zip]
> loading model
> building read index
CPU: 1% (64 cores) * Memory: 6956/254984MB * Storage: 24/841GB * Net: 12↓/0↑MBps
CPU: 0% (64 cores) * Memory: 6981/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7008/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7008/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7003/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7017/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7010/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7006/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7012/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7021/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7011/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7013/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7030/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7082/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7022/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7028/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7027/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7023/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps
CPU: 0% (64 cores) * Memory: 7015/254984MB * Storage: 24/841GB * Net: 0↓/0↑MBps

Any idea why this taking so long? Are there any benchmark available?