Illegal instruction (core dumped) when running PopPUNK #127
Comments
|
Hi,
My feeling is it's the latter, but if we can get pp-sketchlib to compile for you locally we can rule that out. If you still get the error I think it's likely graph-tool's fault, and I can look into that instead. |
|
No problem, thank you for making this tool. I've cloned the master branch for I've tried running the command |
|
Ok, does seem like a sketchlib issue then. Can you run |
|
Running Edit: |
|
Great, thanks. And you are still getting the Illegal Instruction error? Sorry if this is a pain, but would it be possible to try the following:
I would greatly appreciate it if you could help with this, as I've been unable to replicate the problem myself! |
|
I just ran it again (after removing the old I tried running |
|
Ok, so perhaps there was an instruction in the conda version that was unsupported by your CPU. Good that the manual install resolved that – it may have to be the official work-around for this issue. The new error looks like a problem with your h5py installation, the error is just coming from |
|
These are the results from Steps to reproduce (in case one of these steps might cause it to error out): I just noticed that for And here it removes the |
|
So if you run For reference, I have: It might first be instructive to run:
Another suggestion would be that perhaps you could try downgrading to this version of h5py with |
|
Yep, the python in the conda environment has it, but when poppunk tries to run it it gives an error. I noticed that even when though I have h5py through python it did not show up in the conda list. Output from And the output from Running which downgrades pp-sketchlib as well as changing the poppunk version for some reason - to 2.3.0. If I run it with Running this instead then: Where the installation point of hdf5 seems to in |
|
A few notes on the above:
|
|
So after running This is all without messing with Sorry for the many issues! |
|
Ok, hopefully looks like the environment is sorted now! Feels like we're getting there. |
|
I might have spoken too early. I was indeed a stale/corrupt Afterwards running the usual poppunk command I get: with a single file I just checked my own reference_list.txt file - and I forgot that I had modified it and removed some genomes so it wouldn't take as long to test it out. After redoing the |
|
Ah, great! Sorry it was a bit of a long route to get it working! 300 Staph, if you're using a few cores, should really only be of the order of a few minutes. The main computation step is |
|
Absolutely no problem. Thank you so much for helping throughout the whole ordeal! |
|
That's probably having trouble fitting the model in that case. Can I suggest you start with two separate runs, one with (if on poppunk 2.2.0 or lower, the first command becomes It might be worth having a look through the (long) tutorial on the docs to get an idea of how to deal with different fits, but feel free to post some of your output plots here and I can take a look |
|
Sorry for taking some time to go through it! I've tried running it with Network summary:
Components 95
Density 0.0501
Transitivity 0.9941
Score 0.9443The DPGMM graph I got from And the db_scan: From what I can tell the fit still needs to be improved (although the network summary does show good values?). I've tried to go through the tutorial and tune some parameters to see if I can get a better fit, but due to not much experience in it, it has been a bit more difficult for me to understand. Thank you for the all the help! |
|
Both of those fits look ok to me, the DBSCAN fit more specific, the GMM fit broader. I'd say the DBSCAN fit looks best, but I'd run with You could also try running |
|
Closing due as appears to be solved. Please do open another issue, or email me, if you want to discuss your fitting further. |
Hi,
I've been trying to get PopPUNK running on a server to no avail - with the main error being:
PopPUNK (POPulation Partitioning Using Nucleotide Kmers) (with backend: sketchlib v1.5.3 sketchlib: /home/ubuntu/anaconda3/envs/ppunk/lib/python3.8/site-packages/pp_sketchlib.cpython-38-x86_64-linux-gnu.so) Mode: Creating clusters from assemblies (create_db & fit_model) Illegal instruction (core dumped)all of this being on a newly made conda environment
I've seen #106 had a similar problem and tried following the steps that were mentioned there, but none of them have solved the issue.
Looking at the
apport.logfile from the dumped core it only states:I've tried compiling pp-sketchlib from scratch as I've read that might cause some problems (especially regarding the CPU) but it presents its own problem when trying to build it.
Running
python3 setup.py installin the cloned pp-sketchlib repository yields:This is all with the command:
poppunk --easy-run --r-files reference_list.txt --output lm_example --threads 8 --plot-fit 5 --min-k 13 --full-dbRunning this locally on my computer does not seem to give the same error on the other hand when following all the same steps - so I'm stumped for now on what the problem could be
The text was updated successfully, but these errors were encountered: