I just went to use nvcc for the first time and this nonsense hit my firewall. Make won’t compile but it has to do with my unwillingness to use the proprietary toolkit. This network activity only happened once on startup.
I just went to use nvcc for the first time and this nonsense hit my firewall. Make won’t compile but it has to do with my unwillingness to use the proprietary toolkit. This network activity only happened once on startup.
I have to admit I’m not entirely convinced these requests are coming from a compiler…
Is it possible for you to virtualize an non-networked system with your GPU passed through? That seems like the best option IMO. Next best thing would be to set up an airgapped machine just for this, but not everyone has a 2nd machine.
Personally when I was trying out local LLMs I used a virtual machine, mainly due to the known code execution vulnerability related to Tensorflow model data being saved in python’s Pickle format. I believe the recommended save method changed twice since those days though.
With the Firefox stuff I’m assuming you’ve also checked their enterprise config options too (which are admittedly difficult to find and piece together online), so not too sure what else you could do except aside from continuing to block them. Librewolf however provides documentation on how to disable the extra requests sent from their browser in their FAQ, and covers some preferences on the subsequent page linked at the bottom
Normally, I would be quite skeptical of what could be involved, and indeed my ability to diagnose the cause is limited. It is somewhat speculative to draw a conclusion. However, the machine is always behind this whitelist firewall, the only new software on the system was the llama.cpp repo and nvcc, and I’ve never encountered a similar connection anomaly.
I tried to somewhat containerize AI at first, but the software like Oobabooga Textgen defeated this in their build scripts. I had trouble with some kind of weird issue related to text generation and alignment. I think it is due to sampling but could be due to some kind of caching persistence from pytorch? I’ve never been able to track down a changing file so the latter is unlikely.
I typically only use regular FF for a couple of things, including Lemmy occasionally. Most of the extra nonsense on the log is from regular FF. Librewolf is setup to flush everything and store nothing. It only does a few portal checks an hour for whatever reason. I should look into stopping it. With regular FF I just don’t care or use it for much of anything. I just haven’t blocked it in DNF.