amkozlov / raxml-ng

RAxML Next Generation: faster, easier-to-use and more flexible
GNU Affero General Public License v3.0
379 stars 64 forks source link

Tolerance Failing #73

Open Brendanjlk10 opened 4 years ago

Brendanjlk10 commented 4 years ago

While the program is running, with command /home/mrsn/Programs/raxml-ng_v0.9.0_linux_x86_64/raxml-ng --all --msa O167.raxml.rba --model GTR+G --prefix O167 --seed 211476 --threads 3 --bs-metric fbp,tbe --tree pars{100},rand{100} --bs-trees 1000 the following error message appears: Raxml-ng_Error_Message The number of raxml-ng: /home/... error messages corresponds to number of threads given. Number of tree searches performed before failing varies. Changing seed effects that but doesn't solve it. Rerunning the command keeps the progress of the previous run. Also get the error when running the command /home/mrsn/Programs/raxml-ng_v0.9.0_linux_x86_64/raxml-ng --all --msa O167.raxml.rba --prefix O167 --seed 9513 --threads 2 --bs-metric fbp,tbe --tree pars{100},rand{100} --bs-trees 1000 The program does finish eventually, I just have to run the command many times and change the seed each time.

amkozlov commented 4 years ago

I recently fixed a very similar issue in the dev branch, could you please check if it works on your dataset:

https://github.com/amkozlov/raxml-ng/wiki/Installation#building-development-branch

Thanks!

Brendanjlk10 commented 4 years ago

That fixed the issue for me. Though, in testing it, I did find that the program crashes (at least for me in Ubuntu 18.04.3 LTS running Raxml-NG off of data from Panseq with a 251.8 GiB RAM and 56 CPU threads) when I set --blmin too low (for me 1e-10 or 1e-11 and lower) when invoked like: /home/mrsn/Programs/raxml-ng-dev/bin/raxml-ng --all --msa O168.raxml.rba --model GTR+G --prefix O168v13 --threads 7 --blmin 1e-11 --bs-metric fbp,tbe --tree pars{50},rand{50} --bs-trees 150 as seen in the below screenshots of two different errors from two different datasets of highly related isolates. The_Raxml-ng_Error_2019-09-05_11-23-44 The_Raxml-ng_Error_2019-09-05_11-06-25

amkozlov commented 4 years ago

Glad to hear that the original issue was fixed! As for the other error, I'm not surprised that extremely low brlens cause problems. I can have a closer look if you'll send me your dataset, Maybe one thing to try would be to reduce optimization range by adjusting --blmax.