cryosparcm installation seems successful, but I get the following error in the end of cryosparcw installation. I tried doing a reinstall of cryosparc, but get the same results like this:
Attempting to register worker exx2-US-Desktop-Aegis-RS to command exx2-US-Desktop-Aegis-RS:39002
Connecting as unix user exx2
Will register using ssh string: exx2@exx2-US-Desktop-Aegis-RS
If this is incorrect, you should re-run this command with the flag --sshstr
Connected to master.
Current connected workers:
Autodetecting available GPUs…
Traceback (most recent call last):
File “bin/connect.py”, line 221, in
gpu_devidxs = check_gpus()
File “bin/connect.py”, line 91, in check_gpus
num_devs = print_gpu_list()
File “bin/connect.py”, line 23, in print_gpu_list
import pycuda.driver as cudrv
ModuleNotFoundError: No module named ‘pycuda’
What OS and version are you running? Which version of cryoSPARC are you trying to install? Do you have CUDA installed, and does nvidia-smi output GPU information? This seems familiar. If you scroll up the console output from the install, do you have errors during the install of pycuda?
I encountered a similar sounding issue recently, but some more information will be helpful in diagnosis.
Thanks very much for your help. I am installing the newest version of cryoSPARC. I also have CUDA installed. And yes, I do have errors during the install of pycuda. Please see the detailed information here:
Attempting to register worker exx2-US-Desktop-Aegis-RS to command exx2-US-Desktop-Aegis-RS:39002
Connecting as unix user exx2
Will register using ssh string: exx2@exx2-US-Desktop-Aegis-RS
If this is incorrect, you should re-run this command with the flag --sshstr
Connected to master.
Current connected workers:
Autodetecting available GPUs…
Traceback (most recent call last):
File “bin/connect.py”, line 221, in
gpu_devidxs = check_gpus()
File “bin/connect.py”, line 91, in check_gpus
num_devs = print_gpu_list()
File “bin/connect.py”, line 23, in print_gpu_list
import pycuda.driver as cudrv
ModuleNotFoundError: No module named ‘pycuda’
Is the OS Ubuntu? 20.04, 22.04? RedHat derivative?
I had a similar sounding pycuda error on Arch. No guarantee this will help or fix it, but the worst that happens is you delete the install directory and try again…
cd into the cryosparc_worker/bin directory, and run eval $(./cryosparcw env) to enter the work environment, then conda install pycuda -c conda-forge
If that installs correctly, then ./cryosparcw connect --worker [hostname] --master [hostname] --[otherFlags, e.g. --nossd]. Then start cryoSPARC and if there are no errors, please run through the T20S workflow to make sure everything is working.
@rbs_sci
Oh, seems the pycuda install is working as you suggested!
It’s Ubuntu 22.10 OS.
This works for me in this step: cd into the cryosparc_worker/bin directory, and run eval $(./cryosparcw env) to enter the work environment, then conda install pycuda -c conda-forge
Then ./cryosparcw connect --worker [hostname] --master [hostname] --[otherFlags, e.g. --nossd]. showed some errors.
Because now I am not sure of the hostname, I tried various hostnames or no host name like below but all didn’t quite work somehow:
You absolutely need the hostname of both the master and worker nodes for it to work, or it can’t connect them. hostname should tell you the hostname of the box you’re logged in to. If the master and worker nodes are different, you’ll need to find out both separately.
@rbs_sci
I typed hostname in the terminal and find the hostname: exx2-US-Desktop-Aegis-RS, I tried this hostname like below but still didn’t quite work somehow:
The complicated hostname with the hyphens and mixed case in might be causing a problem. Not sure. Never used such a long hostname before. I try to use a single text string, all lowercase.
If I remember systemd syntax correctly, hostname can be changed with: sudo hostnamectl set -hostname newNameHere and rebooting (if using Ubuntu 20.04, for example)…