Hello new project,

I see you have a single windows application available. I was wondering if you plan to branch out to other platforms like Linux.

Also, since this is doing model training and AI model training in general performs better on GPU, any plans for GPU applications, CUDA?

I see that we have a CUDA app now, which is great. And with 1.31, the tasks seem to be completing successfully. But are they actually doing any work on the GPU? I am not seeing any load on the GPUs at all.

waiting for Linux CUDA app.

it's possible the app isnt even CUDA. sometimes projects treat "CUDA" as synonymous with "Nvidia", as Primegrid does.

---

The app is using more than 1 thread but is assigned 1 thread by BOINC.
On a 32 threaded system I see it regularly using 10-15% of the CPU per task.
Memory usage is very high 5-6gb so far (and still climbing I think) which 99% of systems cannot allocate for a single threaded app.

Boinc Manager gives me the following warning on all participating hosts:

"axiom: Data Required to Participate
Important: You must have data in your contribute folder to meaningfully participate in Axiom. Without data, tasks run but submit zero gradients with no impact on the model. Please ensure your contribute folder is set up before expecting meaningful results"

Did I miss anything here? I just joined the project and that was it. Couldn't find any info about getting data and setting up folders...

https://axiom.heliex.net/result.php?resultid=264600

Remove the time limits

Run time 4 hours 10 min 44 sec
CPU time 13 hours 48 min 43 sec

axiom 1/30/2026 12:35:29 PM Output file wu_1_g32203_0_r1589228810_0 for task wu_1_g32203_0 absent

axiom 1/30/2026 6:28:18 PM [error] Error reported by file upload server: Server is out of disk space

The tasks uploads are only 65k.

At top of server status page
Notice: fwrite(): Write of 1360 bytes failed with errno=28 No space left on device in /opt/axiom_boinc/html/inc/cache.inc on line 270

Today new Windows app and a first linux app