Ok, I've got the 7th beta up now (see previous beta 6 thread for URLs). I've changed the algorithm again as outlined in the enclosed readme. Note there's now an extra file, sspred.val which you need.
You may now overwrite a previous beta with this one and you do NOT need to erase filelist.txt.
It SHOULD upload the buffered work, if any, giving you credit as usual if within 48hrs of now, and then it should begin fresh from gen. 0. The generation you are currently working on will be discarded however (thus no credit for partially complete generations, sorry).
Let me know if this is how it appears to work or if you suspect a bug. It should also clean up all old files (provided they are in filelist.txt) properly and not leave them behind from the old protein (unless they were already orphaned to begin with).
We are excited to see the results from this trial as it is significantly different from the previous betas as the gen. 0 'seed' structure is being chosen in a totally new and seemingly better way (and without knowledge of the native structure!) We may still try un-doing the change from beta 5 to beta 6 since that change seemed to worsen results, but again, we will only change one thing at a time.
Good luck
Howard Feldman
Hm.
Howard, would be nice to test this new version. I took the files and wrote them over the beta 6's files.
...
Handle found. Using rod9saoi. To change, hit 'C' and wait 5 seconds...[foldtrajlite] FATAL ERROR: [001.008] Cannot open rotamer library
Hit Return pascal@linux:~/Projekte/df_aktuell>
...
So..I need a full version of the beta client. Can someone give me the link or mail the beta 6 to me?
p.p.k@wanken.de
(I'm waiting).
Thx a lot.
I assume this is why Gen 0 is taking ~twice as long to generate than Beta 6...hopefully it works as it should and produce much better resultsOriginally posted by Brian the Fist
We are excited to see the results from this trial as it is significantly different from the previous betas as the gen. 0 'seed' structure is being chosen in a totally new and seemingly better way (and without knowledge of the native structure!) We may still try un-doing the change from beta 5 to beta 6 since that change seemed to worsen results, but again, we will only change one thing at a time.
Good luck
All you need for running Beta 7 is to load your subdirectory (folder, for the MicroSofties) with the standard version. Then overlay the 5 files for Beta 7 from the following:
Linux version FTP
Windows text client FTP
Windows screensaver FTP
As always, Windows folders can get dfGUI for Beta here .
Enjoy!
HOME: A physical construct for keeping rain off your computers.
Are the RMS values meant to be so high..all my clients are in the 40s & 50s This is embarrassing
From the readme -Originally posted by Grumpy
Are the RMS values meant to be so high..all my clients are in the 40s & 50s This is embarrassing
I assume the lower the score the better (currently sitting at 41.9ishnote gen. 0 'RMSD' is no longer RMSD but a fitness score)
Doh, always helps to look at the readme....
The update ran OK, and considering the trouble I had with one Client over the past 24 hours, this is a good thing
Howard, they are bombing out all over the place.
Using w2k, beta6 was rock solid.
Extracted the beta7.zip file overwrote the files in the DFbeta directory where the beta6 was working, various motherboards/video cards.
========================[ Apr 19, 2003 5:05 PM ]========================
ERROR: [001.009] {foldtrajlite2.c, line 4679} Error occured while aligning structures
ERROR: [001.023] {rotate.c, line 641} Invalid PMMD given to GetRMSD
ERROR: [001.010] {foldtrajlite2.c, line 4683} Abnormally small RMSD: 0.000000 (157 1 1)
FATAL ERROR: [157.096] {foldtrajlite2.c, line 4697} Invalid/corrupt native structure, please download a fresh copy of the software
Getting that as well - can't get further than Gen 0...
Code:========================[ Apr 20, 2003 1:31 AM ]======================== ERROR: [001.009] {foldtrajlite2.c, line 4679} Error occured while aligning structures ERROR: [001.023] {rotate.c, line 641} Invalid PMMD given to GetRMSD ERROR: [001.010] {foldtrajlite2.c, line 4683} Abnormally small RMSD: 0.000000 (157 1 1) FATAL ERROR: [157.096] {foldtrajlite2.c, line 4697} Invalid/corrupt native structure, please download a fresh copy of the software
Last edited by pfb; 04-19-2003 at 08:38 PM.
Ooooops! Yup, same here. Dead in the water.
What was I saying earlier about the meaning of "beta"?
Howard, any files you need posted / emailed?
HOME: A physical construct for keeping rain off your computers.
Yep - it's lucky we have the main client to fall back onOriginally posted by Paratima
What was I saying earlier about the meaning of "beta"?
Its working fine on the PC I updated that is using the Windows text client, so this bug doesn't affect everyone.
A member of TSF http://teamstirfry.net/
what gen are you on? and how did you perform the update?Originally posted by Aegion
Its working fine on the PC I updated that is using the Windows text client, so this bug doesn't affect everyone.
Then you're the lucky one, Aegion. I got a W2K and a W98 box, both hung.
I just fired up the Beta text client. No problems to report as of struct 5000 gen 0 (other than it took me a while to figure out the difference between registering at beta. and the regular site).
===bob briggs
OOOOPSSSSS. I seem to have put the native.val file from the 'live' version in the beta zip packages. SORRY
The version Im testing on my home machine here is still on gen. 0 so I didn't realize yet
If you still have the native.val from beta 6, that one will work fine. I'll fix up the update packages on the FTP site right now.
Howard Feldman
Just let us know when it's OK.
Also, will we need to delete any old files?
Letting us know when it's fixed by posting something on the news page would be good. Page hasn't been updated since January
===bob briggs
All 3 of my clients are now on Gen 1 and going fine.....
Hah..just checked all 3 error logs...they are not as fine as I thought...
Let us know when the fixed versioin is up
It sounds like you are missing the rotamer library., rotlib.bin.bz2. This is not included with the beta but rather comes with the standard package available at www.distributedfolding.org/Download.htmlOriginally posted by Pascal
Hm.
Howard, would be nice to test this new version. I took the files and wrote them over the beta 6's files.
...
Handle found. Using rod9saoi. To change, hit 'C' and wait 5 seconds...[foldtrajlite] FATAL ERROR: [001.008] Cannot open rotamer library
Hit Return pascal@linux:~/Projekte/df_aktuell>
...
So..I need a full version of the beta client. Can someone give me the link or mail the beta 6 to me?
p.p.k@wanken.de
(I'm waiting).
Thx a lot.
Possibly you deleted it and anyhow it has nothing to do with not having tested the program (and I don't appreciate the implication either
Howard Feldman
all seems ok now - just overwrote the old native.val with the new one...
on a different note - is the beta website down? I can't seem to get it to show (keeps saying it's unavailable)...
edit - more info: the domain name cannot be resolved... nor is it uploading
edit #2 - not uploading due to 'previous generation missing' - still generates folds though...going to restart from gen 0 and see what happens....
this is after 'Q'uitting and replacing native.val...Code:========================[ Apr 20, 2003 2:24 AM ]======================== ERROR: [000.000] {foldtrajlite2.c, line 3861} Cannot find structure from previous generation ; find it manually or delete filelist.txt to continue ERROR: [000.000] {foldtrajlite2.c, line 4019} Error during upload: Previous generation missing
Last edited by pfb; 04-19-2003 at 09:27 PM.
The beta website HTML text is not being updated. It is merely there for your convenience. Anyways the files are fixed (but make sure your browser doesn't cache!). I've attached native.val as a ZIP archive to this message to for convenience.Originally posted by rsbriggs
Letting us know when it's fixed by posting something on the news page would be good. Page hasn't been updated since January
===bob briggs
Just overwrite the incorrect one. The only consequences are you will probably need to quit ('Q') and restart because it reads in the file at startup I believe. The native structure is not used for gen. 0 so thats why the error doesnt show up immediately.
Howard Feldman
Just a note: 'Q'uitting didn't do it for me. Both instances had to be forcibly disincarnated.
Grabbed the file from the attachment. Thanks!
(I was a bit concerned about the file posted on the website - the name "distribfold-current-win9x" rather than "distribfold-beta-win9x" - thought maybe it was the "release" version rather than the beta version).
I probably won't finish up the last 4500 in gen 0 for another 1/2 hour, so I won't know anything until that rollover happens....
===bob briggs
same here (eventually) - only figure after gen 3 that it wasn't uploading so deleted filelist and it seems ok(ish)...Originally posted by IronBits
I had to delete my filelist.txt file, after replacing the native.val file.
No biggy, it's just getting started anyways I guess.
Does it normally upload after each generation?
Would deleting the filelist file and re-starting be a good idea?
So the Beta download zip file now has the correct Val in it ?
Yep - when I re-downloaded itOriginally posted by Grumpy
So the Beta download zip file now has the correct Val in it ?
I deleted all my filelist.txt just to be sure. Before I did that, they were all on Gen 1 or 2 with RMS of 17-19, back to 54 now, hope this is OK. This is why we are here Folding with a Beta..we do not expect things to run smooth , that would be too boring
rsbriggs -- yes, it does normally upload after each generation (unless you have it configured not to, with "-i f" in the foldit script).
Anyone -- Is it taking a relatively high CPU percentage for anyone else when your client stops to calculate the energy? When this beta does that, it starts the calculation by doing something that takes about 50% CPU, kernel time, for 3-4 seconds. Then it finishes in user time, but normal priority. The other beta did this too, but it did it between generations instead (it may have also done it during gen 0; I wasn't paying any attention then).
In case I haven't explained it well enough (probably the case), I've attached a gkrellm screenshot illustrating the effect. The green is normal, nice'd time, and the orange spots are what I'm talking about. Time flows left-to-right.
I have four nodes crunching offline (-if). I stopped them, copied in the new files, and restarted them.
I then ran my upload script, which stops the offline client, then runs the client with -ut, then restarts the offline client.
All four nodes failed to upload due to a "Previous generation missing" error.
I've now removed the old work manually (after making a backup of one of the directories).
From the error.log:
========================[ Apr 19, 2003 10:06 PM ]========================
ERROR: [000.000] {foldtrajlite2.c, line 3861} Cannot find structure from previous generation ; find it manually or delete filelist.txt to continue
ERROR: [000.000] {foldtrajlite2.c, line 4019} Error during upload: Previous generation missing
The work files in the directory (note: the *_0_* ones appeared when I started the beta7 client):
<handle>_0_protein_0001994.val
<handle>_1_<handle>_protein_238_0000042_min.val.bz2
<handle>_1_<handle>_protein_239_0000050.val
<handle>_1_<handle>_protein_239_0000050_min.val
<handle>_protein_240.trj
fold_0_<handle>_1990_protein.log.bz2
fold_1_<handle>_45_<handle>_protein_239.log.bz2
The contents of the filelist.txt:
./fold_1_<handle>_40_<handle>_protein_238.log.bz2
./<handle>_1_<handle>_protein_238_0000042.val
fold_1_<handle>_45_<handle>_protein_239.log.bz2
<handle>_1_<handle>_protein_239_0000050.val
fold_0_<handle>_1990_protein.log.bz2
<handle>_0_protein_0001994.val
CurrentStruc 0 2245 125 0 0 1994 49.649 -1265.738 1459.233 10.700 137427824.000 0.850 1.500 250.000 ----------------------HHHHH-HHHH----HHHHH---------------HHHHHHH---------------------------------
1e5d1d5970e68fa9fe40f0cd4aae04f7
Hmmmm, strange things are happening on the Stats page. It says my lowest RMSD is 13.77, yet underneath for the Top 10 RMS it lists mine at 41.18...please explain :bs:
UPDATE: Doh, the Stats pages are just slow at catching up to what has been logged..very strange that the left hand of the Server has no idea what the right is doing, I mean the top of the Stats say 1 figure and the bottom of the page another.:
Note: Has the way data is uploaded and processed by the Server changed with this Beta, or has it always had a big delay between logging your best RMS and transferring that to the Top 10 RMS
Last edited by Grumpy; 04-20-2003 at 01:01 AM.
Several different things on the Beta stats page get updated at different intervals. It'll all make more sense when this becomes the main client, and the data is updated at the same interval.
Howard (the duck.. erm.. Doc
Well, up to Gen 17, and no errors in the logs of my Dualie....it is quiet, too quiet :sleepy:
I am not a Stats Ho, it is just more satisfying to see that my numbers are better than yours.
The stats pages do seem slow :
I have a smallest RMSD structure of 9.38 (gen 19 according to dfGUI - nearly finishing 20 at the moment) yet the top 10 shows me at #2 with 10.14 (gen 15)...
How slow is the update...vaguely remember it being 1/2 hour rather than 10 mins - which is odd as the last update time is 03:01...just under 50 minutes ago...
Gen 21 and all is quiet here. Program is spending the vast majority of its time in the "TIGHT SPOT - TRYING ALTERNATIVE CONFORMATION #79423" mode.
Couldn't you list the Best Energy So Far right on the console screen? Hate having to keep opening up progress.txt to see how things are going.....
===bob briggs
FreeDC Mercenary
Thanks a lot for your tipps. I didn't setup the beta 7 the right way. Now it's working fine.
I was just wondering if it would be possible to add in a scaling factor when the beta becomes the main client for people running older machines.
Considering that it took pretty much everybody 8-10 days to go through a full set (250 gens) with fairly powerful machines, I think that people with older machines may not be able to complete a full set before a change over takes place.
Perhaps they could have their generation sizes scaled down by a variable factor (2,5, 10?). I realize this may affect the quality of the results received, however, what's preferable, someone getting work for the full size through gen 25 or getting work for 1/10th the size through gen 250?
Just something to think about.
Team Anandtech DF!
Posting Permissions
Reply With Quote
