thulle
Sighted testing and different audio levels are other common sources of perceived vinyl superiority.
Many people enjoy the ritual of grabbing a vinyl and setting it up for playback, having physical access to coverart and maybe lyrics, which are entirely valid reasons to prefer vinyl, but that isn't audio quality.
> This approach enables Heretic to work completely automatically. Heretic finds high-quality abliteration parameters by co-minimizing the number of refusals and the KL divergence from the original model. This results in a decensored model that retains as much of the original model's intelligence as possible. Using Heretic does not require an understanding of transformer internals. In fact, anyone who knows how to run a command-line program can use Heretic to decensor language models.
Abliteration seems to be what Heretic does?
> Now these "abliterated" models all suffer from catastrophic breakage because they are not as simple anymore.
I'm not knowledgeable about each step in the process of making these abliterated models, but some more popular ones with steps after Heretic, seem to improve on the benchmarks tried of the base model:
arc/c arc/e boolq hswag obkqa piqa wino
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF [instruct mode]
mxfp8 0.711,0.879,0.910,0.790,0.514,0.823,0.763
mxfp4 0.701,0.873,0.909,0.786,0.488,0.813,0.759
Qwen3.6-27B-Instruct: [base, non heretic]
mxfp8 0.647,0.803,0.910,0.773,0.450,0.806,0.742
I'm not seeing any similar benchmarks of the HauhauCS models, at least the ones I checked, so I assumed the opinion is based on your own trials, but then you argue in favour Heretic. Is the based on pre-Heretic abliteration techniques? Which might then not be appliable to this "Proprietary weight-level abliteration developed by the dealignai research team."?
A comment points out that they aren't randomized:
> The values used were chosen such that it takes a large number of bit flips to change from allowed to denied. Using random values doesn't really protect against this attack.
Nitter: I get all 8 posts in the thread in 18 requests, 207 kB, 169 kB transferred.
X: I only get the first post, 128 requests, 11.45 MB, 2.11 MB transferred.
Assuming they're not selling the 4090 at a loss and being generous that the whole cost is the chip cost, that makes for a $1600 chip. Scaled to the H100 we get $2140. That's a pretty minuscule fraction of a $35k card.
That doesn't say anything about the fraction of the cost on the finished product though? Those numbers just say it could be something like 2% vs 6% of the cost of the card.
Output generated in 101.74 seconds (0.98 tokens/s, 100 tokens, context 82, seed 532878022)
Output generated in 515.46 seconds (0.99 tokens/s, 511 tokens, context 27, seed 660997525)
Checking nvidia-smi it stalls at ~130W (out of ~470 W max) power usage, ~25% GPU usage and ~10% memory bandwidth usage. There's fairly much traffic on the pci-bus though, and the python process is stable at 100% usage of one core. GPU possibly limited by some thing handled in python? Pausing the GPU-accelerated video-decoding of a twitch stream it get a surprisingly large boost:
Output generated in 380.42 seconds (1.34 tokens/s, 511 tokens, context 26, seed 648992918)
They'd need 20 months to manage to notify users?
Holding off on 2.2 seems recommended, and if you're keeping critical data on OpenZFS it might be a good idea to give the issues a glance. The 2nd one might have the same underlying solution as an issue that has given me system freezes when closing in on ~90% pool usage using 2.1 on top of LUKS.
That might not've been a consideration when the offending posts were removed. The ones doing the removal might not even be the same people handling suspensions.
> I could find nothing even questionable the banned person did.
That might be an issue of legitimacy. I'm involved in some discord modding, and we got admin-only channels mirroring the others to keep logs of everything. I have no idea how suspensions are handled in the NixOS community though, maybe moderators are picked to handle things in their own discretion.
> The fact that you don't agree with some person on an issue is no grounds for a ban.
That's not what's happening here though, it's the intentionally provocative behaviour?
Say some person argues that Israel is a nazi state dehumanizingly enough to get a warning, and gets told to stop derailing everything into a discussion about that since it's a detriment to a community that in itself has no connection to the issue. Agreeing to this they instead start sporting a green flag, or a little image with a free country between a river and a sea - maybe with a green flag on it - while linking to a page about how Israel is a nazi state. Then the issue the moderators have to consider isn't whether they agree on how to describe Israel, but if the actions of the warned person are still acting inflammatory and derailing the community.
A bit over the top example to make it clearer, if you find it an incomparable situation just tone down the actions in it accordingly.
Not sure you can link to them then?
Considering the contents of the page seem to support their conclusion that the image was just another way to intentionally provoke others in the community I guess they figured the banned member had no intention of following the intention of the warning, but skirting it as close as possible. And now when that failed and they got banned they're instead trying to get people to rally to their support.
I have no clue how fast the temperature equalizes or how long it takes to scorch milk at different temperatures, but I heard about this as a kid and made the following thought experiment.
If I pour the milk first and then put in a single drop of tea, the milk at the surface of the drop would receive some flash heating, then equalize pretty fast to an average of their temperatures and then as the water turbulences away the resulting temperature would barely be above that of the original milk.
If I pour the tea first and put in a single drop of milk, the relative masses would equalize the temperature of the milk to pretty close to boiling, and it's gonna require a lot of drops before the milk cools down again.
Thus the time spent at higher temperature seems longer in the latter case. Assuming it matters for things like protein breakdown, and that those changes can affect the taste, then the question becomes: how slowly do we have to pour our tea for this to matter? I suspect I don't pour anywhere near slow enough.
Also 2014 seems to've been more ice due to record snowfall, which is easier to freeze than the regular freezing of the sea. Seems to be in line with climate models predicting more erratic weather. Good example of why they weren't worried about that, since it's expected, but worried about this, since it isn't.
Can you expand on this? If it's geoengineering you're talking about we don't really need all the energy for it.
If we just reduced the resources needed to produce todays consumption, why wouldn't the savings just go to producing more of the same stuff? Ie. growth?