I'm not sure whether this fix should go in openmmtools.mcmc or here in YANK, but to avoid the issue where velocities are left at zero when sampler states are restored from disk, we should reassign velocities when restoring sampler states. Otherwise we see this behavior:
I'm not sure whether this fix should go in
openmmtools.mcmc
or here in YANK, but to avoid the issue where velocities are left at zero when sampler states are restored from disk, we should reassign velocities when restoring sampler states. Otherwise we see this behavior: