A community effort to collect a curated set of analysis pipelines built using Nextflow.

nf-core

^Regarding this, what is the max mem we can give a module for the github runners? I found the max is 15GB on the modules repo, but folddisco seems to preload its feature space in memory no matter the input, which requires 16.4GB. How to tackle?

<https://runs-on.com/docs/runners/labels/#ram> to make it pass once and then reverse it.

Revert in the same PR after it gets the green light, and then you force push like usual? I should leave a comment in the code with the minimum viable memory for the module also I guess

<@U05LNHCFLCW> did you try with different Hash algorithms as well with `--type` . According to claude, `pdb` has the lowest mem footprint on RAM.

I am doing that now!!! `-y pdb` fixes that :wink:

```     args       │              peak RSS              │
├──────────────────┼────────────────────────────────────┤
│ default (-t 4)   │ 16.4 GB ← 2³² slots × 4 B = 16 GiB │
├──────────────────┼────────────────────────────────────┤
│ -t 1             │ 16.4 GB (threads irrelevant)       │
├──────────────────┼────────────────────────────────────┤
│ -d 8 -a 3        │ 16.4 GB (bins irrelevant)          │
├──────────────────┼────────────────────────────────────┤
│ -y 3di           │ 8.2 GB                             │
├──────────────────┼────────────────────────────────────┤
│ -y pdb -d 8 -a 3 │ 1.0 GB                             │
├──────────────────┼────────────────────────────────────┤
│ -y trrosetta     │ OOM-killed even here```

This ur own benchmarking, if so how did u do it so fast?

agents had already done it :stuck_out_tongue: