Hi all,
I didn't get an answer when I asked before about this, but I thought that I would try again. In DMRG the common practice is to use the truncation errors at different MAXSTATES to extrapolate the energy. However, the dmrg code currently does not write out the truncation error to a file. Is there a way to have the dmrg program write this out for the last sweep? Even just one value from the last sweep or the mean would be useful. I know that the truncation error for each step is written out to the screen, but it is rather inefficient to run one batch at a time and then copy down the truncation error. I really appreciate the help that I have received thus far and am very grateful for the ALPS collaboration overall. I'd appreciate any hints you might have regarding this.
Thanks, Justin Peel
Justin,
I had no trouble getting the truncation error when I ran batch jobs. Just put stdout in a file, then do the appropriate grep command to find the end of each MAXSTATES calculation and get the truncation error you want. Something like 'grep "LAST SWEEP" -B30 | grep truncation' should give you what you want. You might also try debug printing, but your output and stdout will be much larger.
If the details I just described don't work, it's because I haven't run Alps DMRG in three months and have forgotten the exact syntax of the correct commands, but when I was running it, I wanted exactly the same information you do, and it didn't take much to get it.
If my suggestion isn't satisfactory even in principle, you can also add a judicious print statement (perhaps with logic to limit it to the final iteration) to the source code and run that way. I found Adrian's code very easy to understand and I'm not even a C++ person.
Best,
Jeff
On 9/15/07, Justin David Peel justin.peel@utah.edu wrote:
Hi all,
I didn't get an answer when I asked before about this, but I thought that I would try again. In DMRG the common practice is to use the truncation errors at different MAXSTATES to extrapolate the energy. However, the dmrg code currently does not write out the truncation error to a file. Is there a way to have the dmrg program write this out for the last sweep? Even just one value from the last sweep or the mean would be useful. I know that the truncation error for each step is written out to the screen, but it is rather inefficient to run one batch at a time and then copy down the truncation error. I really appreciate the help that I have received thus far and am very grateful for the ALPS collaboration overall. I'd appreciate any hints you might have regarding this.
Thanks, Justin Peel
We have added the output to the current svn version, and it will be in the next release which will happen as soon as I have investigated one other bug report.
Matthias
On 16 Sep 2007, at 07:23, Jeff Hammond wrote:
Justin,
I had no trouble getting the truncation error when I ran batch jobs. Just put stdout in a file, then do the appropriate grep command to find the end of each MAXSTATES calculation and get the truncation error you want. Something like 'grep "LAST SWEEP" -B30 | grep truncation' should give you what you want. You might also try debug printing, but your output and stdout will be much larger.
If the details I just described don't work, it's because I haven't run Alps DMRG in three months and have forgotten the exact syntax of the correct commands, but when I was running it, I wanted exactly the same information you do, and it didn't take much to get it.
If my suggestion isn't satisfactory even in principle, you can also add a judicious print statement (perhaps with logic to limit it to the final iteration) to the source code and run that way. I found Adrian's code very easy to understand and I'm not even a C++ person.
Best,
Jeff
On 9/15/07, Justin David Peel justin.peel@utah.edu wrote:
Hi all,
I didn't get an answer when I asked before about this, but I thought that I would try again. In DMRG the common practice is to use the truncation errors at different MAXSTATES to extrapolate the energy. However, the dmrg code currently does not write out the truncation error to a file. Is there a way to have the dmrg program write this out for the last sweep? Even just one value from the last sweep or the mean would be useful. I know that the truncation error for each step is written out to the screen, but it is rather inefficient to run one batch at a time and then copy down the truncation error. I really appreciate the help that I have received thus far and am very grateful for the ALPS collaboration overall. I'd appreciate any hints you might have regarding this.
Thanks, Justin Peel
-- Jeff Hammond The University of Chicago
comp-phys-alps-users@lists.phys.ethz.ch