Hello, I’m documenting this in case it helps others,
This turned out to be a misleading error caused by incompatible CPU instructions. On my system, the CPU supports AVX but not AVX2. Gubbins ends up calling raxmlHPC-AVX2 under the hood, and the later failure is reported only as:
Unable to fit model to data
So the issue is not the alignment itself, but that an AVX2-optimized RAxML binary is being selected on a host that does not support AVX2.
It would be a nice quality-of-life improvement if a future release could detect CPU compatibility earlier and fail with a more explicit message before the run gets this far.
Command example
run_gubbins.py \
--threads 1 \
--starting-tree fasttree_phylogeny.tre \
--model GTRCAT --recon-model GTRCAT \
filtered_alignment.fasFor context, I'm currently using Gubbins v3.3.5 from this container depot.galaxyproject.org-singularity-gubbins-3.3.5--py39pl5321he4a0461_0.img
Hello, I’m documenting this in case it helps others,
This turned out to be a misleading error caused by incompatible CPU instructions. On my system, the CPU supports AVX but not AVX2. Gubbins ends up calling
raxmlHPC-AVX2under the hood, and the later failure is reported only as:Unable to fit model to dataSo the issue is not the alignment itself, but that an AVX2-optimized RAxML binary is being selected on a host that does not support AVX2.
It would be a nice quality-of-life improvement if a future release could detect CPU compatibility earlier and fail with a more explicit message before the run gets this far.
Command example
run_gubbins.py \ --threads 1 \ --starting-tree fasttree_phylogeny.tre \ --model GTRCAT --recon-model GTRCAT \ filtered_alignment.fasFor context, I'm currently using Gubbins v3.3.5 from this container
depot.galaxyproject.org-singularity-gubbins-3.3.5--py39pl5321he4a0461_0.img