More model version explanations

[callum-b]

New issue checks

• I did not find an existing documentation enhancement.

Documentation change request

I am quite new to Dorado and ONT in general. I am following the docs to run basecalling via the NextFlow pipeline, and when I use the help option, I am greeted with this:

I managed to find the docs on Dorado that explain what the names mean, but I'd like more information on them. When I try to run the nf basecaller pipeline demo dataset, I am told that I don't have the 32GB of required RAM, but only 31.2 GB. Surely there is some other model I could use that is slightly lighter?

I was unable to find out this information, or where I could find this info. A handy guide of RAM usage / time per kb / accuracy for each model would be very handy to better estimate what software to run on my hardware.

If this information is available somewhere, I was unable to find it, and it would be nice to be directed towards it in the "Models" section of the Dorado docs.

[malton-ont]

Hi @callum-b,

That limitation is enforced by the workflow, not by dorado - you'd need to discuss that on their repo.

Regarding RAM usage, this is generally not the limiting factor for dorado as basecalling is performed by the GPU. For CPU basecalling, dorado will scale the number of runners to the available GPU memory. This may be affected by additional processing such as alignment or barcoding.

For accuracy information, please see here.

We don't publish benchmark data for each model and release at present, but there's a helpful community effort here.