Message boards : RALPH@home bug list : Rosetta mini beta and/or android 3.61-3.83
Previous · 1 . . . 7 · 8 · 9 · 10 · 11 · 12 · Next
Author | Message |
---|---|
zioriga Send message Joined: 16 Feb 06 Posts: 8 Credit: 323,279 RAC: 1,175 |
Now I'm at 99.5 after 31,5h On my S3 |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
Now I'm at 99.5 after 31,5h These are wus for I7 or Ryzen, not for my little arm... :-( |
dekim Volunteer moderator Project administrator Project developer Project scientist Send message Joined: 20 Jan 06 Posts: 250 Credit: 543,579 RAC: 0 |
I agree. I'll talk to the researcher. The long run time is due to filtering logic. Even on one of our local machines it takes about an hour to generate a model (after quite a few filtered attempts). |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
I've another one. 99,1% after 18h. 99,6 after 37h....a small agony Now, i think 1000 points is earned!!! :-P |
Vikram K. Mulligan Send message Joined: 6 Jan 16 Posts: 2 Credit: 0 RAC: 0 |
Although these slow jobs take longer to find a solution, they are heavily biased to sampling regions of conformation space that these peptides can actually access and which are relevant to our interests. That is, we might make the sampling rate slower by a factor of 100, but we're perhaps 1,000 or 10,000 times more likely for any given sample to be "useful". We're crunching numbers more intensively to produce better results instead of just relying on more sampling. I can say that the jobs that we're running now are successfully sampling the native state with fewer samples in cases in which incredible amounts of the old, naïve sampling produced nothing near the native state. I'm not sure that we can make the sampling faster (i.e. produce more samples per CPU-hour), but the credit earned per sample should be proportional to the CPU-hours spent per sample, no? |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
We're crunching numbers more intensively to produce better results instead of just relying on more sampling. This is not a problem, i like the science you do!! the credit earned per sample should be proportional to the CPU-hours spent per sample, no? I'm not sure. If i crunch a decoy in 2h with, for example, a Ryzen 1800x and the same decoy with my Huawei in 30h, i take the same points? |
dekim Volunteer moderator Project administrator Project developer Project scientist Send message Joined: 20 Jan 06 Posts: 250 Credit: 543,579 RAC: 0 |
Since the android app is a separate application, the credit per decoy running average will only use results from android jobs. But there is the issue of possibly adding a correction factor to try to normalize the credits coming from the standard app vs the arm app given the same protocol. I'll have to look into this and it would require running some benchmarks and updating the standard app. The arm app is running the latest Rosetta version. |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
99,6 after 37h....a small agony Now 99,7% after 57h. Strange behaviour, it continues to slow down. I don't think i'll finish before deadline (tomorrow morning). |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
Although these slow jobs take longer to find a solution, they are heavily biased to sampling regions of conformation space that these peptides can actually access and which are relevant to our interests. i'm afraid, but it's 99,8% after 105h. I've killed this wus. |
Admpicard999 Send message Joined: 4 Sep 16 Posts: 2 Credit: 1,257 RAC: 0 |
Hi: I see there are many here who have tasks that have been running for quite some time. I have three that have gone well past their reply by time: one running for 74 hours at 99.2%, a second 29 hours at 98.1% and a third at 30 hors at 98.2%. This is using a Pixel on Android 7.1.2 (newest). Do I let them keep going or do I stop them? |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
Do I let them keep going or do I stop them? I know these working units are important for admins/developers, but i kill all "ralphtest_c." wus All with these warnings: WARNING: linker: ../../projects/ralph.bakerlab.org/minirosetta_android_3.83_arm-android-linux-gnu: unused DT entry: type 0x6ffffffe arg 0x2828 |
dekim Volunteer moderator Project administrator Project developer Project scientist Send message Joined: 20 Jan 06 Posts: 250 Credit: 543,579 RAC: 0 |
please keep those going. |
dekim Volunteer moderator Project administrator Project developer Project scientist Send message Joined: 20 Jan 06 Posts: 250 Credit: 543,579 RAC: 0 |
You can ignore those warnings. They are harmless. I will hopefully have those warnings removed in the next release soon. |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
please keep those going. 50h to crunch 0,1% i think 2 options: - a bug in "ralph_test_c." wus - my smartphone is too little to crunch these wus and i need an S8 I'll try another wu |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
I just finished correctly a "ralp_test_c." wu after 20k seconds. But these wus are different from problematic "ralphtest_c."?? |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
- a bug in "ralph_test_c." wus "ralphtest_c.", i mean |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
|
zioriga Send message Joined: 16 Feb 06 Posts: 8 Credit: 323,279 RAC: 1,175 |
I have this problem: In my "Tasks for user" page I have 4 WU in progress, but ....... on my smartphone I have 5 WUs. :-) I don't know if this is a problem specific to android app or is a general problem |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
99,9% after 111h of crunching. Still 99,9% after 145 hours. What should i do? |
mmonnin Send message Joined: 6 Dec 16 Posts: 10 Credit: 392,615 RAC: 0 |
I aborted those that I had at 99%. The others just timed out while the server was down. |
Message boards :
RALPH@home bug list :
Rosetta mini beta and/or android 3.61-3.83
©2024 University of Washington
http://www.bakerlab.org