| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: Work unit availability |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 822
|
|
| Author |
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: 
|
To get opng jobs and run a specific other wcg CPU project (only it)? I can't find it. The workaround doesn't work so good, so I'm asking. That was asked for earlier (by me, among others). Apparently it requires some high-level discussion and re-configuring at WCG. I haven't heard back about it since. It is an obvious thing to want, but as long as they have enough GPUs to support the project, I doubt they will make it a high priority. |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
Are there a different amount of work units for AMD and NVIDIA in the pipeline? It seems like most people are using NVIDIA cards. I am thinking of getting a new GPU and i wonder what will be the most beneficial if work units are limited. thanks When a work unit goes on the feeder, it gets assigned to 0, which means first type of device to try to grab it is allowed to grab it. During the initial release of the application, all devices are claimed as unreliable and need to be compared with a wingman. So, if a wingman is required, it gets put on the feeder with a required app version that prevents all other classes from getting it. Usually we have a mixture of results that need a specific class and others that are completely open. Thanks, -Uplinger |
||
|
|
Richard Haselgrove
Senior Cruncher United Kingdom Joined: Feb 19, 2021 Post Count: 360 Status: Offline Project Badges: 
|
... it gets put on the feeder with a required app version that prevents all other classes from getting it ... I'm a little nervous about that methodology. If there's a subtle but generic fault in one particular variant (and I wrote to you about one such possible case on Monday), 'confirmation bias' will lead to other machines of the same class validating the WU and disguising the error.SETI@Home chose, from the very beginning, to encourage users to build apps from the open sources to run on their particular obscure operating system or hardware. Those tasks were required to take their place in the rough-and-tumble of validation against any other variant of the science application, whether project-built or self-built. Some subtle (but minor) errors were picked up that way. |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges:
|
... it gets put on the feeder with a required app version that prevents all other classes from getting it ... I'm a little nervous about that methodology. If there's a subtle but generic fault in one particular variant (and I wrote to you about one such possible case on Monday), 'confirmation bias' will lead to other machines of the same class validating the WU and disguising the error.SETI@Home chose, from the very beginning, to encourage users to build apps from the open sources to run on their particular obscure operating system or hardware. Those tasks were required to take their place in the rough-and-tumble of validation against any other variant of the science application, whether project-built or self-built. Some subtle (but minor) errors were picked up that way. Hey Richard, I did not get a chance to write back about this kind of issue, but we have additional validation that is done on the backend that is not a bitwise comparison as it sounds like SETI was initially using. We worked with the scientist to come up with a way to validate results that are sent back to us that can use but don't require a wingman for validation. We also spot check all 'reliable' hosts and can mark them as unreliable when they don't pass a single validation. All of this was discussed with the researchers at length and took months to get the validator to where it is now. They also have additional checks on their end during the post analysis. Thanks, -Uplinger |
||
|
|
Richard Haselgrove
Senior Cruncher United Kingdom Joined: Feb 19, 2021 Post Count: 360 Status: Offline Project Badges:
|
No, SETI was not using bitwise validation. Each task returned up to 30 signal candidates: the number/type of signals reported had to match exactly, and for each signal candidate I think four (possibly more for some signal types) mathematical values had to match within a given tolerance. But having drawn your attention to the possibility of a problem, I'll leave it there. We can return to it when you start on v2.0, if you want.
|
||
|
|
Unixchick
Veteran Cruncher Joined: Apr 16, 2020 Post Count: 967 Status: Offline Project Badges: 
|
It is great to see some of my seti friends on the boards. Thanks to the team that made gpu processing possible. I'm enjoying trying them out on my machine.
|
||
|
|
sam6861
Advanced Cruncher Joined: Mar 31, 2020 Post Count: 107 Status: Offline Project Badges:
|
To get opng jobs and run a specific other wcg CPU project (only it)? In website device profile,- Turn on "If there is no work available for the project(s) I have selected above, please send me work from another project." - Turn off OpenPandemics and turn on other projects. - Turn on some or all of: Use AMD / NVidia / intel graphics card. - Make sure project limits is very high or unlimited. (scroll down to the bottom) - If you want to, BOINC project preference, store 0.5 days or less. CPU can always do the selected projects as long as selected projects have some CPU work available. GPU will check if other projects have GPU work, nothing there, then with "please send me work from another project." on, can send GPU tasks from OpenPandemics. There appear to be very little amount of OpenPandemic GPU work so it may take a while to grab any GPU tasks. |
||
|
|
gibbcorp
Advanced Cruncher Joined: Nov 29, 2005 Post Count: 77 Status: Offline Project Badges:
|
Thank you for this. Looking forward to improving my output.
|
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
To get opng jobs and run a specific other wcg CPU project (only it)? In website device profile,- Turn on "If there is no work available for the project(s) I have selected above, please send me work from another project." - Turn off OpenPandemics and turn on other projects. - Turn on some or all of: Use AMD / NVidia / intel graphics card. - Make sure project limits is very high or unlimited. (scroll down to the bottom) - If you want to, BOINC project preference, store 0.5 days or less. CPU can always do the selected projects as long as selected projects have some CPU work available. GPU will check if other projects have GPU work, nothing there, then with "please send me work from another project." on, can send GPU tasks from OpenPandemics. There appear to be very little amount of OpenPandemic GPU work so it may take a while to grab any GPU tasks. Im pretty sure this will not work if he’s using the OPNG profile that has send work for CPU set to no even if he turns off OPNG and enables other projects. He will have to set send CPU work back to yes. Much easier to use 2 distinct profiles and just switch back and forth between them IMHO.
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
Michael Goetz
Cruncher United States Joined: Dec 11, 2017 Post Count: 35 Status: Offline Project Badges: 
|
To get opng jobs and run a specific other wcg CPU project (only it)? I can't find it. The workaround doesn't work so good, so I'm asking. That was asked for earlier (by me, among others). Apparently it requires some high-level discussion and re-configuring at WCG. I haven't heard back about it since. It is an obvious thing to want, but as long as they have enough GPUs to support the project, I doubt they will make it a high priority. YES, there is a way to make this work. It's a little more complicated than setting a few options on a webpage, however. 1) Create TWO BOINC installations on your computer. You will need to set a special BOINC option to make this work. Include <allow_multiple_clients>1</allow_multiple_clients> in the <options> section of cc_config.xml, and then install two copies of BOINC in different directories with different data directories. Failure to properly set <allow_multiple_clients> will usually result in the server declaring most of your tasks as errors. 2) Create two venues at WCG -- one set up to run only GPU tasks, and one set up to run only CPU tasks. 3) Attach both BOINC installations to WCG, and assign one to the first venue and the other to the second venue. Alternate method: 1) Create two venues, as before. One for the CPU apps, one for the GPU apps. 2) Attach your BOINC installation to the WCG and use the GPU venue. 3) Create a VM on your computer, and inside the VM install BOINC. Attach this copy of BOINC to WCG and have it use the CPU venue. The host machine's BOINC will run the GPU tasks, and the BOINC inside the VM will run the CPU tasks. Note that you can not reverse the venues; GPU apps can't run inside the VM. This second option only makes sense if you already have a VM running on the computer. For example, a few of my Windows machines have Linux VMs so I can run Linux-only when needed. |
||
|
|
|