What version of Guppy are you using? My guess is it's an old version that was released prior to your kit being released.

$ singularity exec --nv ONT_Guppy_GPU.sif guppy_basecaller --version
Guppy Basecalling Software, (C) Oxford Nanopore Technologies, Limited. Version 6.0.1+652ffd1

$ singularity exec --nv ONT_Guppy_GPU.sif guppy_basecaller --print_workflows
flowcell       kit               barcoding config_name                    model version
FLO-FLG111     SQK-RBK110-96     included  dna_r10.3_450bps_hac           2021-04-20_dna_r10.3_minion_promethion_384_72309afc

As an aside, even with a small MinION run, 2 callers with 1 thread per caller will take a very long time to run. It would be worthwhile trying to get a GPU build and a supported GPU if at all possible. Run times go from dozens of hours to a few minutes from experience.

If your guppy build is not GPU friendly, then the parameter -x cuda:all should be redundant since they refer to GPU cores.

Finally, since your kit is a barcoded multiplexed kit, I'm assuming you want to demultiplex the samples too. I have no experience with this, but it appears you are missing the demultiplexing parameter. Check out the ONT guide here.

EDIT: Just realised you said the error is it is not a GPU build. I thought that was a statement about it. You'll need a membership account with ONT to download and build a new version of the GPU caller. It looks like you have an old CPU only version.

To use GPU, you probably need a GPU card with at least 4-6 GB GPU RAM. Which one do you have ?

I put some example commands which I used to use (on A100 GPUs with 40 GB RAM) here:

https://github.com/colindaven/guppy_on_slurm/blob/master/runbatch_gpu_guppy.sh

Also, make sure you get the GPU guppy version from the ONT community as dthorbur says.