I realised today that dorado duplex is giving me the following error on pretty much every basecalling job I submit. I'm splitting the pod5 by channel and parallelising across several A100 GPUs on an HPC environment. I'm using dorado 0.7.1 with v5 models and modification calling. The jobs are not dying thought but I'm concerned about whether some of the data might be corrupted. It would be good to know whether I should wait for a fix or keep basecalling.
Hi again,
I realised today that
dorado duplex
is giving me the following error on pretty much every basecalling job I submit. I'm splitting the pod5 by channel and parallelising across several A100 GPUs on an HPC environment. I'm using dorado 0.7.1 with v5 models and modification calling. The jobs are not dying thought but I'm concerned about whether some of the data might be corrupted. It would be good to know whether I should wait for a fix or keep basecalling.Some do end up seg faulting though: