I've had an issue with getting ribodetector to work in Apptainer/Singularity (using latest pipeline release 3.27.0 or prior). I'm aware of a previous issue with ONNX multiprocessing hang, that had to do with previous ribodetector versions but supposedly this has been fixed for ribodetector 0.3.3. Even when adjusting '--chunk_size' to smaller values, running the pipeline with gpu ribodetector always tries to allocate around ~ 75GB to my GPU when I only have ~50GB available. here is part of my config: // withName: '.*RIBODETECTOR' { // queue = 'gpu_h200' // clusterOptions = '--gres=gpu:1' // containerOptions = { workflow.containerEngine == 'singularity' || workflow.containerEngine == 'apptainer' ? '--nv' : // ( workflow.containerEngine == 'docker' ? '--gpus all' : null ) } // memory = { 120.GB * task.attempt } // cpus = 8 // time = 24.h // // // This prevents the H200 from hitting a CUDA OOM // ext.args = '-l 99 -m 180 --chunk_size 4096' // } For the cpu version of ribodetector, it just goes on forever and never finishes (which makes no sense). Should I try to allocate more GPU? sortmerna works fine in every run, which I have defaulted to because ribodetector has never worked for me.

Could you file a bug at github.com/nf-core/rnaseq/issues please, with as much information as possible including all relevant logs? Thanks!