Posts by Yifan Song

1) Message boards : Number crunching : Minirosetta 3.46 (Message 75596) Posted 10 May 2013 by Yifan Song Post: Thanks! I'll let the person running the idealdead2 jobs know about this. -y
2) Message boards : Number crunching : Minirosetta 3.46 (Message 75581) Posted 7 May 2013 by Yifan Song Post: I've been running debugging from my side for the last week on the same set of jobs, it's running a lot slower with the debug mode, so I haven't consistently reproduce the seg fault yet. My suspicion is that something is still not quite fixed in the gradient calculations.
3) Message boards : Number crunching : Minirosetta 3.46 (Message 75545) Posted 30 Apr 2013 by Yifan Song Post: OK, I got it wrong earlier. The Windows/x86 version is for the actual application, not the graphics. For some reason that file didn't get updated the last time I ran the script. I just reran the update, and it looks ok now. I didn't even think that would be the problem. Sorry about the confusion.
4) Message boards : Number crunching : Minirosetta 3.46 (Message 75535) Posted 29 Apr 2013 by Yifan Song Post: rb_04_26_38593_73094__t000__0_C1_SAVE_ALL_OUT_IGNORE_THE_REST_79295_1489_0 (task ID 578191167) died with exit status -1 in less than 12sec using the new code. I'm running a local test now, and it's been running for 20 min now and still going. Maybe there's some downloading errors that make the input files incomplete? OK, found the problem with this one. It comes from our robetta server using a parameter to randomly trigger a deprecated function. I just changed the server to disable that mechanism.
5) Message boards : Number crunching : Minirosetta 3.46 (Message 75529) Posted 28 Apr 2013 by Yifan Song Post: Hi Yifan. I had this one finish O.K. but showing this error message. I also have another 3 cryo tasks that are running overtime and have only check pointed once, I'll let them go to see what happens. cryo_bh__chain_K_subrun_002_SAVE_ALL_OUT_IGNORE_THE_REST_79122_932_0 http://boinc.bakerlab.org/rosetta/workunit.php?wuid=524887806 # cpu_run_time_pref: 21600 dof_atom1 atomno= 3 rsd= 1 atom1 atomno= 1 rsd= 1 atom2 atomno= 2 rsd= 1 atom3 atomno= 5 rsd= 1 atom4 atomno= 6 rsd= 1 THETA1 nan THETA3 nan PHI2 0 ERROR: AtomTree::torsion_angle_dof_id: angle range error ERROR:: Exit from: src/core/kinematics/AtomTree.cc line: 780 ====================================================== DONE :: 52 starting structures 21197.3 cpu seconds This process generated 52 decoys from 52 attempts ====================================================== BOINC :: WS_max 0 BOINC :: Watchdog shutting down... BOINC :: BOINC support services shutting down cleanly ... called boinc_finish </stderr_txt> ]]> Validate state Valid That comes from the same problem as the sin_cos_range error. I'm looking into it now. The bug in the gradient calculations makes the transformation matrix non-orthogonal, which is why arcsin(cos) gets the bigger-than-one input, an then some angels become NaN
6) Message boards : Number crunching : Minirosetta 3.46 (Message 75527) Posted 28 Apr 2013 by Yifan Song Post: rb_04_26_38593_73094__t000__0_C1_SAVE_ALL_OUT_IGNORE_THE_REST_79295_1489_0 (task ID 578191167) died with exit status -1 in less than 12sec using the new code. I'm running a local test now, and it's been running for 20 min now and still going. Maybe there's some downloading errors that make the input files incomplete?
7) Message boards : Number crunching : Minirosetta 3.46 (Message 75526) Posted 28 Apr 2013 by Yifan Song Post: CASP9_fb is a really old batch of jobs. The symmetry definition IO changed since then. So the new executable shouldn't work on them any more. "com" defines the center of mass, and I believe the naming was changed to avoid confusion. y This one ran under the new app & had this error message 99 times by the look of it, I haven't counted them ;) you can if you like. CASP9_fb_benchmark_hybridization_run54_T0613_0_D2_SAVE_ALL_OUT_IGNORE_THE_REST_48029_1425_1 http://boinc.bakerlab.org/rosetta/workunit.php?wuid=524562110 ERROR: error in process_residue_request: 'com' ERROR:: Exit from: src/core/conformation/symmetry/util.cc line: 93 ====================================================== DONE :: 99 starting structures 1384.62 cpu seconds This process generated 99 decoys from 99 attempts ====================================================== BOINC :: WS_max 0 BOINC :: Watchdog shutting down... BOINC :: BOINC support services shutting down cleanly ... called boinc_finish
8) Message boards : Number crunching : Minirosetta 3.46 (Message 75513) Posted 27 Apr 2013 by Yifan Song Post: Thanks guys! The cryo jobs use a different protocol, so they do run longer. Let me take a look at the sin_cos_range error. That was the error I eventually saw with the bug in the 3.45 version. I'll check to see if there's anything else still causing the problem. I think the Windows/x86 one is only for graphic interface, the actual minirosetta program runs on the platform "Microsoft Windows running on an AMD x86_64 or Intel EM64T CPU". I'll double check with DEK to make sure. I'll also tell the user running the abinitio job to pay attention to their input files. Yifan PS: the cryo_bf... jobs are from earlier, I think the input files might be been screwed up already with earlier iterations using the old release.
9) Message boards : Number crunching : Problems and Technical Issues with Rosetta@home (Message 75495) Posted 27 Apr 2013 by Yifan Song Post: the memory problem turned out to be tricky. As least for my local test, it wasn't running out of memory, or memory leak. Turned out, there was a problem with the potential gradient calculation in the electron density energy function. And this makes the reference frame to drift away, and eventually crash. I just updated the code and restarted the jobs. If you still see the errors after updating the application, please post here: http://boinc.bakerlab.org/rosetta/forum_thread.php?id=6222 I'll be monitoring that thread closely. Yifan
10) Message boards : Number crunching : Rosetta Application Version Release Log (Message 75494) Posted 26 Apr 2013 by Yifan Song Post: minirosetta is updated to 3.46 with new development and bug fixes. http://boinc.bakerlab.org/rosetta/forum_thread.php?id=6222
11) Message boards : Number crunching : Minirosetta 3.46 (Message 75493) Posted 26 Apr 2013 by Yifan Song Post: minirosetta is updated to 3.46 to include recent developments in electron density and other scoring functions. This update also fixes a bug in density gradient calculations that drives the reference frame apart and occasionally cause the program to crash in a long simulation. Post problems related to the update here.
12) Message boards : Number crunching : Problems and Technical Issues with Rosetta@home (Message 75466) Posted 25 Apr 2013 by Yifan Song Post: Hi, I'm sorry for causing all the trouble with my cryo work units. The crashes are related to using electron density data. I'm updating r@h with bug fixes that should make these jobs more stable. Yifan
13) Message boards : Number crunching : Rosetta Application Version Release Log (Message 68307) Posted 31 Oct 2010 by Yifan Song Post: minirosetta is updated to 2.17 to fix the memory and invalidation errors. http://boinc.bakerlab.org/rosetta/forum_thread.php?id=5502
14) Message boards : Number crunching : minirosetta 2.17 (Message 68306) Posted 31 Oct 2010 by Yifan Song Post: This update fixes the memory and invalidation issues from 2 weeks ago. please report bugs here.
15) Message boards : Number crunching : minirosetta 2.16 (Message 68020) Posted 10 Oct 2010 by Yifan Song Post: My apologies for the new errors. I forgot that those jobs are using an option that's associated with 2.15. I just cancelled those jobs.
16) Message boards : Number crunching : Rosetta Application Version Release Log (Message 68005) Posted 9 Oct 2010 by Yifan Song Post: minirosetta is reverted to 2.14 (now 2.16) http://boinc.bakerlab.org/rosetta/forum_thread.php?id=5475
17) Message boards : Number crunching : minirosetta 2.16 (Message 68004) Posted 9 Oct 2010 by Yifan Song Post: This is reverting minirosetta to 2.14 due to the recent memory problem.
18) Message boards : Number crunching : minirosetta 2.15 (Message 67961) Posted 5 Oct 2010 by Yifan Song Post: After David Kim and TJ looked into this, we did find a problem with large memory usage with the 2.15 version. I'll do a revert first thing tomorrow. (Too tired to get it started now :p)
19) Message boards : Number crunching : minirosetta 2.15 (Message 67892) Posted 1 Oct 2010 by Yifan Song Post: There are quite a few reports here on large memory usage with the latest jobs. Due to the ongoing post-CASP analysis, I think a lot of the jobs are for difficult targets, which means large proteins. And they tend to eat up more memory than normal jobs do. I'll double check with the people submitting these jobs.
20) Message boards : Number crunching : Rosetta Application Version Release Log (Message 67781) Posted 21 Sep 2010 by Yifan Song Post: minirosetta 2.15 is out. see here for the update http://boinc.bakerlab.org/rosetta/forum_thread.php?id=5453

Next 20