Closed f-brinkmann closed 1 year ago
Seems to work fine! I would also output ram.available once in the beginning to double check the configuration, maybe after the ram detected print.
@chris-hld: No including in echoing the parameters max_ram_load: 30.00 GB (31.05 GB detected, 27.68 GB available)
I will wait with the patch release of this until we figured #104
I can report that the RAM resource management now works as expected. I will highlight that at least on my system 30s between call and measuring RAM again was not enough time to fully register/load. With 60s in between this seems resolved. Last commit produced this output
-----------------------------------------------------
Running 13/256 unfinished frequency steps in the project
1/13 starting instance from 'FABIAN' (May 29 2023, 10:43:37)
source 2, step 128, 22050.0 Hz
estimated 81.55 GB RAM of available 380.00 GB required
2/13 starting instance from 'FABIAN' (May 29 2023, 10:44:37)
source 2, step 106, 18260.2 Hz
estimated 56.11 GB RAM of available 302.74 GB required
3/13 starting instance from 'FABIAN' (May 29 2023, 10:45:37)
source 1, step 128, 22050.0 Hz
estimated 81.55 GB RAM of available 249.50 GB required
4/13 starting instance from 'FABIAN' (May 29 2023, 10:46:37)
source 2, step 108, 18604.7 Hz
estimated 59.77 GB RAM of available 172.23 GB required
5/13 starting instance from 'FABIAN' (May 29 2023, 10:47:38)
source 1, step 127, 21877.7 Hz
estimated 81.55 GB RAM of available 115.53 GB required
... waiting for resources and checking every second (May 29 2023, 10:48:38)
5 NumCalc instances running at 25.70% CPU load
41.32 GB RAM available (401.78 GB used), 62.76 GB required
6/13 starting instance from 'FABIAN' (May 29 2023, 12:15:36)
source 2, step 109, 18776.9 Hz
estimated 59.77 GB RAM of available 63.78 GB required
... waiting for resources and checking every second (May 29 2023, 12:16:36)
6 NumCalc instances running at 32.70% CPU load
7.2 GB RAM available (435.89 GB used), 66.73 GB required
7/13 starting instance from 'FABIAN' (May 29 2023, 12:50:50)
source 2, step 126, 21705.5 Hz
estimated 78.95 GB RAM of available 79.71 GB required
... waiting for resources and checking every second (May 29 2023, 12:51:51)
6 NumCalc instances running at 13.90% CPU load
25.52 GB RAM available (417.58 GB used), 66.73 GB required
8/13 starting instance from 'FABIAN' (May 29 2023, 13:00:02)
source 1, step 112, 19293.8 Hz
estimated 63.55 GB RAM of available 100.36 GB required
... waiting for resources and checking every second (May 29 2023, 13:01:02)
7 NumCalc instances running at 14.10% CPU load
40.16 GB RAM available (402.93 GB used), 72.47 GB required
9/13 starting instance from 'FABIAN' (May 29 2023, 13:20:59)
source 2, step 122, 21016.4 Hz
estimated 73.07 GB RAM of available 73.85 GB required
... waiting for resources and checking every second (May 29 2023, 13:21:59)
6 NumCalc instances running at 14.20% CPU load
24.96 GB RAM available (418.14 GB used), 72.47 GB required
Hi, I have re-installed my system and currently do not even have Python installed yet. + going on a long vacation soon, So I will not be of much help during the summer.
Here are the changes in the RAM management https://github.com/Any2HRTF/Mesh2HRTF/pull/101/commits/79d07c09236bce8aad6c768331dc7df481822c66
And this is an example of the new output
@chris-hld, this should be pretty much identical to #100 but does not use
psutil.psutil.virtual_memory().used
because the documentation mentions that this is for informal purposes only. It seems to work well on my machine. Can you check it on the cluster the next days?