Definitely, I had also done a lot of investigation and assumed that when I isolated the problem to the RAM there was almost nothing more I could do expect cry and wait for the inevitable torturous moment I’d have to open my wallet amid sky high prices.

The server has now been running non-stop without an issue for 30 days, so I assume the problem was fixed by the suggestion!

BERT is a special case. The academic consensus is rightly considering it an LLM since it operates on text but the industry doesn’t consider it an LLM since LLM in its colloquial meaning has drifted to mean a model generating text (which BERT doesn’t). Whisper is not an LLM since it’s working on audio, so it is an Automatic Speech Recognition model (ASR).

Text in = LLM in academic circles (BERT and GPT)

Text in, text out = LLM in all circles (GPT only)

Audio in, text out ≠ LLM (Whisper)

Now, we can argue whether we want to accept the colloquial meaning of LLM, but the fact is that Whisper is not an LLM. And neither ASR nor BERT causes even a fraction of the damage GPT is doing, but that’s another discussion.

LLM is a subset of NLP. All LLMs are NLP but not all NLP are LLMs. To be more precise, LLM is a method of achieving NLP while NLP is the goal. Whisper is just a specialised speech-to-text engine. It doesn’t have the text reasoning engine that is the core of LLMs. Yes, it has a dataset to inform it’s encoder and decoder, but that doesn’t make it an LLM. If you strip out the encoder, then you’re getting closer.

Update, kernel panic happened today, so I updated the BIOS settings:

• tCL (CAS Latency): Change from 11 to 12
• tRCD: Change from 11 to 12
• tRP: Change from 11 to 12
• tRAS: Change from 28 to 32
• DRAM voltage changed from 1.488 to 1.530

All the initial values were Auto values. I’ll let it sit for at least a month with no reboots before calling it fixed. I’ll set a reminder

I’ll try to remember to update you in 1-2 months then!

All the documentation/readme is AI generated, so I guess the difference is AI. I can’t speak for the code itself since I’m not a TS expert but almost the entire code is dumped in an initial commit which I definitely wouldn’t do, even when building the first prototype. But that’s not a dead giveaway.

For a project I liked I saw that the translation of the UI was only about 1% complete for my native language. I decided to dedicate the weekend to get it above 95%. This was apparently the part that seemed too hard for most to make a dent in, but across several new versions my language is still 100%. I think my groundwork made it easier for some sporadic contributors to get a small bit of dopamine when updating the few new strings in a release to see it hit 100% again.

I’ll try that next time it dies, see if it helps

The server currently runs on 32gig but I was planning on upgrading to 64 or 128gig DDR5. Could go as low as 16 if need be, but the the price is still hurting more than having to reboot 2-4 times a month.

Only thing I wish for Christmas is to be able to replace the RAM and mobo in one of my servers. It is failing with a kernel panic about once a week and it is an old DDR3 platform I planned to upgrade back in January. I have isolated the issue to a hardware fault but left without a solution…

I was more thinking of being able to “mark” part of the 24/7 recording as a drive to work so it could be found and aggregated later.

You know, it’s free to self host, so I’ll just experiment and see how it works :)

Looks great!

Small question, is it possible to log a specific journey? For example in Denmark you are required to keep a precise log of your driving if you want to get a tax rebate, so if you could log a specific journey and save it for later for tax documentation, that would be great.

I speak gibberish.

It is a framework that allows different AI programs to coordinate their tasks, provide sources for their conclusions, and verify each other’s work without relying on a single central server. It should help avoid drift when solving complex problems.

I kinda liked posting some random funny snippets of TV shows and movies on the forbidden Lemmy. Maybe I should take that up again now that Cliparr can cut down most of the time it’ll take.

Only problem is that I still need the communities that would benefit from such submissions. On Reddit this was mostly on a fan sub for the given TV show, but Lemmy isn’t big enough for that (at least I think so).

About $3/mo. But for a lifetime deal you’re also buying the risk. If they go bankrupt, stop honoring the lifetime deal, or any variation thereof tomorrow, you’re out $750 - lifetime deals, where they exist are often heavily discounted compared to normal rates due to this. 20 years is though quite a long time. Plex is only 16 years old.

In a perfect world a company would limit the amount of lifetime deals available and only have them in the beginning to get some quick cash allowing them to scale. I don’t think Plex is running a very good business, which also devalues the lifetime deal.

who prefer a subscription for a tracker

You’re in the wrong tracker if that is the case. No trustworthy tracker has a subscription.

That’s simply not true. The Decathlon HRM Band for example connects to virtually everything that can understand a generic HRM over BLE. And limited in what sense? It’s either sending a heart rate or not.

OpenTracks has integrations with several HR BLE trackers: https://codeberg.org/OpenTracksApp/OpenTracks/src/branch/main/README_TESTED_SENSORS.md

OpenTracks?