Iván Pulido
Iván Pulido
Related https://github.com/choderalab/perses/issues/800
As discussed, it could be very useful and not too hard to have a way to specify the number of processes for each of the phases. As a subkey-value pair...
By using `openmm-nightly` and `LangevinDynamicsMove` we are getting almost a 4x speedup for the neutral test:  This means that for the neutral mutation test, for our target of 3000...
> Great! Also, I just checked and the [current tests](https://github.com/choderalab/perses/blob/main/perses/tests/test_repex.py) use 3000 iterations -- Why are you using 1000 iterations? Is the test still passing (i.e. `assert DDG < 6...
> Ah that comment was referring to the charge changing mutations. For the neutral mutations, you'll have to see my comment [here](https://github.com/choderalab/perses/issues/1044#issuecomment-1165575916) I see, well I guess we can use...
@zhang-ivy Ah good catch, I just checked and I did use a wrong node in one of the tests (I misread `ls-gpu` nodes on Lilac to be 1080ti and these...
> This would be responsible for a 3x difference, but it's hard to imagine the 1080Ti vs 1080 is responsible for the remaining remaining ~2.7x iteration time difference. Yeah, that...
> @ijpulidos : Can you try re-running all the tests? And if you are still seeing an 8-fold difference, let's set up a quick chat to discuss what might be...
> However, I don't have something to compare it to where I serialize out the state and reload it in a different process to see it get faster by >2x....
I don't have a strong opinion here, there are valid points from all sides. I guess that as long as we make this clear from the docstrings and existing documentation,...