QMCPACK
A wrong DMC calculation
In the process of DMC calculations, the following warning appeared, but the task did not stop, ultimately the DMC energy results are wrong.
Yes, if this warning is hit repeatedly the run will likely be biased. An occasional warning is often OK if you have enough additional statistical samples to dilute a small bias at a few steps. The warning is printed when the maximum number of walkers per rank is hit. The computed limit is printed in a line " Max Walkers per MPI rank " at the start of DMC, and obviously 2501 here. You can change it by specifying max_walkers as an input parameter in the DMC section. My general recommendation would be to either not set it at all, or to set it to a giant number, unless you truly need it -- some users have used it to avoid running out of memory when they are running close to memory limits.
If you have not already, check that you have a sensible trial wavefunction, timesteps for DMC are reasonable, and that the DMC equilibration looks fine. (For the latter make a graph of the dmc.dat and check the energy decays as expected and total population stabilizes.) These are very common reasons for misbehaving DMC runs that can result in the population becoming much larger than the target value or to display large fluctuations.
As a practical note, you can inspect the trace plots made with qmca -t ./*dmc.dat to get an idea if hitting the ceiling has impacted your runs. Look for large fluctuations or rapid increases in the number of walkers and inspect the total energy and trial energy in these regions to spot obvious biases.
Please can you post images of these trace plots here?
With your small energy/variance ratio in VMC, it is surprising that you hit the ceiling. What kind of system are you looking at? What is the target walker count?
