ramess101
commented
6 years ago @msoroush
Sorry for the slow response. I finally got around to testing the RecalcFrame branch. The energy recalculation appears to work but I have a few questions/concerns. In brief:
- PDB files work for now, but are not ideal
- Recomputing energies works, but for some reason the energies are not output to the his.dat file
- How can I restart (not recompute) a simulation without needing another equilibration period?
The PDB files will become enormous (around 30 GB for hexane) if we are outputting the configurations at the same frequency that we used to output the histograms (every 200 MC steps). To work around this, I perform hundreds of smaller GCMC simulations sequentially. After each GCMC run, I perform the RecalcFrame using the PDB file for this short run, and then I restart the GCMC run and remove the old PDB file:
- Run first simulation
- Rerun (recompute the energies)
- Restart from previous simulation (load final configuration)
- Loop through steps 2 and 3 until finishes total number of steps
Obviously, this is just a work around, and is hopefully only temporary. Once the configurations are stored in a .trr (Gromacs trajectory) file we should not have this storage issue. Any idea how long until that will be ready?
More importantly, I have a question about RecalcFrame. I need to have the number of molecules and the recomputed energy for each frame in the PDB file. Obviously, N does not change upon recalculation and I could extract the recomputed energies from the log file, but I thought it would be easier to get both from the his.dat file. However, for some reason his.dat is empty for the recalculation. I tried setting SampleFreq to 1 (since I need it to output to his.dat for every PDB frame) but this did not solve the problem. When recomputing the energies, does it no longer output to the his.dat file? I did notice that the block average file outputs the energies at the same frequency as the PDB coordinate frequency. I cannot use block averages for generating basis functions, but it appears that the block averages reported for the recalculation are actually from a single PDB frame. So for now I am using the block average file, although it is somewhat misleading to call these block averages (Note that you must set BlockAverageFreq to 1, because the code divides the single frame values by BlockAveragesFreq.)
Also, I am having some issues with the restarts. Specifically, I have not been able to get the restarts to run without including an equilibration period. I was hoping that if the previous run had already equilibrated and adjusted the move sizes, then the restart would have some way of loading those parameters. Is that not the case? I tried setting EqSteps to 0, hoping that it would just skip the equilibration because it already loaded the move sizes, but this did not seem to work. I do not have a lot of experience using GOMC, so perhaps I made a simple mistake for restarting simulations.
To help debug my problem, I have included the .conf files for the start, restart, and rerun (recalculate). (Note that I tried setting PRNG to RESTART in in_restart.conf, but the simulations would not run so I changed this back to RANDOM).
Thanks
msoroush
YounesN
We implement the total energy recalculation from outputted PDB file by GOMC. We might change console output a little bit but I think it is done. It is implemented in the branched called "RecalcFrame" https://github.com/GOMC-WSU/GOMC/tree/RecalcFrame. You can use it for all Ensembles.
To set it up;
1- set the "RunSteps" to "0".
2- Load the output PDB "OutputName_BOX_0.pdb" and PSF "OutputName_merged.psf".
3- If you need to calculate the pressure, use "PressureCalc true 10000".