jaimergp

Might be related to the CUDA version detection logic. What happens if you set `CONDA_OVERRIDE_CUDA=11 conda info`?

Yes, that's an annoying but safe workaround. Just make sure that env var is set to the version of your CUDA runtime (11.7, if I am reading correctly). The problem...

I think if you run this code, which isolates the buggy part, you can reproduce the errors: ``` $ conda activate base $ python -v -c "from conda.common.cuda import cuda_detect;...

As expected, the last line is the `ctypes` import: ``` import 'ctypes' # ``` And we don't see anything after (like Python cleaning up tasks), so that's when it crashes....

Hi! Would you consider it to be stable enough right now? Otherwise, how can I help? `ase` has it implemented as well -- maybe that's good enough for a benchmark?

I have been testing it against ASE. Maybe I am doing something wrong, but I got ASE's to work and I am not able with YANK's/openmmtools'. Is this a good...

Thanks for your answer! Yes, I can send the files via email. Can you send me your address? In the meantime, I will check the average forces, but I'd say...

> @hannahbrucemacdonald: > [[email protected]](mailto:[email protected]) > > Hmm ok maybe it's not the forces then, it's just the only difference I spotted at first look. If you send me over the...

From the usability point of view, one of the comments I have received in my previous projects is that the amount of files generated by the software was too high....

I will write here some of the ideas we got after our talk with @Olllom. There are performance issues while resuming calculations. All of the replicas accessing the monolithic NetCDF...