2D Classification: Waiting because resources are not available (RAM)

Hi,
I have freshly installed cryosparc v2.8.3 on a ubutnu standalone workstation with cuda 10.1. Everything seems to run fine, except that I can not start a class_2D job due to “Waiting because resources are not available (RAM)”. I read here (2d class jobs queued) about this issue and the problem of having too less RAM installed appears to be true for my setup (12 GB RAM) so I tried adding CRYOSPARC_RAM_SLOTS=“X” to config.sh with X=4 or greater without success.

Is there anything else to try?

Best
BJ

PS: Here is the output of bin/cryosparcm cli "get_config_var(name='targets', colname='sched_config')" :

[{u’lane’: u’default’, u’name’: u’AX370’, u’title’: u’Worker node AX370’, u’resource_slots’: {u’GPU’: [0], u’RAM’: [0, 1], u’CPU’: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]}, u’hostname’: u’AX370’, u’worker_bin_path’: u’/home/…/cryosparc2_worker/bin/cryosparcw’, u’cache_path’: u’/home/…/’, u’cache_quota_mb’: 5000, u’resource_fixed’: {u’SSD’: True}, u’cache_reserve_mb’: 1000, u’type’: u’node’, u’ssh_str’: u’…@AX370’, u’desc’: None}]

Hi @junglas,

You need at least 16GB of RAM to run a 2D classification job. Each ‘ram slot’ in the resource slots array represents 8GB.

Thanks for clarification. I checked my system setup again and I have actually 16 GB of RAM installed (2x8 GB), sorry for the confusion…
So 2D classification needs at least more than 16 GB?
As described here:" All new jobs get queued, don't run " export CRYOSPARC_RAM_SLOTS=“X” should overwrite the RAM-check, or did I get that wrong?

Best
BJ

Hi @junglas,

The link you attached only applies to cryoSPARC v0, which we no longer support. The software architecture between v0 and v2 are completely different. Also, my apologies, I was wrong about the minimum RAM requirements of the 2D Classification job. It actually requires 24GB of RAM.

Nevermind. Meanwhile I figured out how to get it working for me.

I could lower the limit in the RAM-check to 16 GB by editing build.py of the class2D job (and the other jobs that required more than 16 GB RAM). I just changed the value in line 93 from 24000 to 16000:

job.set_resources_needed(2, params.get(‘compute_num_gpus’, 1), 16000, params.get(‘compute_use_ssd’, True))

This worked out fine for. I could go through the tutorial without errors and even processed a set of >100000 particles.

Best
BJ