What is Hackers' Pub?

Hackers' Pub is a place for software engineers to share their knowledge and experience with each other. It's also an ActivityPub-enabled social network, so you can follow your favorite hackers in the fediverse and get their latest posts in your feed.

0

이번에 대선에서 민주당 쪽에서 국민통합 어쩌고 약파는 샛퀴들은 다 걸러내야 함. 박쥐같은 샛퀴들. 지들 당대표 그렇게 갈굼 당할 땐 모른척하더니 이제 처 기어나오는 거 바라. 기회주의자 샛퀴들

0
0
0
0

I need help naming a package!

TLDR:

"Transform a Markdown document so that selected parts of it are transformed via a pipeline that allows the execution of side-effects."

Examples:

- Graphviz. Render graphic for every `dot` code block and insert the resulting image in MD.
- LaTeX. Use PDF/SVG output to preview mathematical formulae (and other LaTeX documents and snippets).
- Table formulae. Re-render Markdown tables by running formulae on cells.

1) md-transform?
2) md-process?
3) ???

0
0
0
0
0
0
0
0

윤씨 탄핵됐다고 슬퍼하기에 다음 대선때 빨간당 뽑지말라고 하니까 그럼 뽑을 사람이 없는데... 해서 그럼 차라리 투표하러 가지 마 했더니 그래도 투표는 해야한대... 아시겠습니까 여러분 우리는 이렇게 부지런한 사람이랑 싸워야합니다

0
0
0
0

가정환경

자식한테 무관심한 가정에서 자라면
그냥 알아서 잘 지내겠지 하고 냅두는 게 무서워진다.
그러면서 내가 자기들 기준에서 벗어나는 직업을 가지려고 하면 죽일듯이 반대하니까
가족들의 관심을 바라는 나는 그 일을 끝까지 붙들고 있을 수가 없다.
그래서 내가 원하는 게 뭔지도 모르게 되고 우울과 무기력에 빠진다.

결국 어른이 되어서까지 영향을 받는 건
나한테 신경쓰지 않아도 된다고 여겨지는 게 무서워진다.
나한테 관심이 없어도 다른 사람들이 관심 있어주겠지 하는 생각에
결국 아무도 관심이 없고 궁금해하지 않을까봐.
나한테 무슨 일이 생겨도 아무런 관심도 없을까봐.

그래서 너무 억울하고 그 때 억울했던 걸
보상받으려는 심리가 생긴다고 해도
내가 원하는 감정이 생기면
일단 그 감정을 의심하지 않고
남들이 뭐라던 그냥 밀고 나가는 게 중요하다.
가족들의 관심이 필요할 나이가 지났으니까.
나 혼자서도 충분히 존재할 수 있고
가족들이 관심이 없어도
나한테 정말 관심이 있는 다른 사람들과 사귀며
그렇게 지내다보면 내 마음의 결핍도 나아질까 싶다.

0
0

Hello, hachyderm! we've been working hard on building up our ansible runbooks and improving hachyderm's overall resilience. Recently, we've been focusing on is database resilience.

We're getting close to retiring our original database server (finally!) and preparing to move to a fully ansible-managed set of databases servers, primary and replica on new hardware. We'll send another announcement when we do the cut over. The team has done excellent work to make this highly automated, quick, and painless! :blobfoxscience:

Done:

✅ author ansible roles for managing postgresql, pgbackrest (backups), pgbouncer, and primary/replica failover
✅ decide to continue with pgbouncer and *not* use pgcat
✅ rotate database passwords
✅ order new replica database hardware
✅ order new future primary database hardware

To do soon:

🟨 rebuild replica database with ansible scripts
🟨 prepare primary database with ansible scripts
🟨 start replicating to new database replica
🟨 cut over to new database server 🎉

We're also planning on open-sourcing our ansible roles in the coming weeks - just a little housekeeping & tidying up before we do!

0

Auf dem kommenden Easterhegg werde ich einen Workshop über die Mandelbrotmenge halten (und auch noch andere Fraktale erwähnen. Ich sammle gerade weiterführende Links für Leute, die da tiefer einsteigen wollen. Hat eins von Euch empfehlenswerte einstiegsfreundliche Erklär-Sachen (Videos, Webseiten, Computerspiele) zu dem Themenbereich?

:BoostOK:

0
0
0
0

Hello, hachyderm! we've been working hard on building up our ansible runbooks and improving hachyderm's overall resilience. Recently, we've been focusing on is database resilience.

We're getting close to retiring our original database server (finally!) and preparing to move to a fully ansible-managed set of databases servers, primary and replica on new hardware. We'll send another announcement when we do the cut over. The team has done excellent work to make this highly automated, quick, and painless! :blobfoxscience:

Done:

✅ author ansible roles for managing postgresql, pgbackrest (backups), pgbouncer, and primary/replica failover
✅ decide to continue with pgbouncer and *not* use pgcat
✅ rotate database passwords
✅ order new replica database hardware
✅ order new future primary database hardware

To do soon:

🟨 rebuild replica database with ansible scripts
🟨 prepare primary database with ansible scripts
🟨 start replicating to new database replica
🟨 cut over to new database server 🎉

We're also planning on open-sourcing our ansible roles in the coming weeks - just a little housekeeping & tidying up before we do!

0
0
0
0
0
0

capitalists have spun the deskilling of labor with AI as democratizing it, as if those that worked hard for their skill unfairly hold it over those without those skills, yet another case of capitalists trying to pit labor against itself

Real democratization of labor would be giving all a chance to learn those skills

0
0

Habt ihr ein Konto bei ? Am 21. April tritt dort eine neue Datenschutzerklärung in Kraft. Die Verkaufsplattform will ab dann eure persönliche Daten für KI-Training verwenden.

Wenn ihr das nicht möchtet, könnt ihr auf accountsettings.ebay.de/ai-pre widersprechen. Dort setzt ihr (nach eurem Login!) den aktivierten Schalter auf "Nein".

Warum es problematisch werden kann, wenn eure Daten in künstliche Intelligenz einfließen, lest ihr auf unserer Internetseite: verbraucherzentrale.nrw/node/1

Tablet auf dem Schoß einer Person zeigt das Logo "ebay" auf dem Display. Dazu Text: "Persönliche Daten für KI: Widerspruch bei Ebay bis 21. April. So geht's!"
0
0
0
0
0
0

어제 갑자기 몸 안좋아진거 있어서
토리키조쿠 오늘 안갈려다가 왔거든?

집에서 밥먹으려고 냉장고에서 반찬 꺼내는데
라쿠텐카드에서 5000포인트가 들어오더라?

다음 주 까지 쓰라는거야
못참지

0
0
0

2025/04/04 21:59:00

USD: 1449.00 ▼ 3.20 (-0.22%)
JPY: 995.23 ▼ 2.13 (-0.21%)
EUR: 1596.15 ▼10.93 (-0.68%)
CNY: 198.87 ▼ 0.60 (-0.30%)

GBP: 1882.98 ▼20.35 (-1.07%)
CAD: 1019.67 ▼11.43 (-1.11%)
HKD: 186.41 ▼ 0.31 (-0.17%)
AUD: 885.70 ▼33.25 (-3.62%)

0
0
0

DeepSeek: Inference-Time Scaling for Generalist Reward Modeling
L: arxiv.org/abs/2504.02495
C: news.ycombinator.com/item?id=4
posted on 2025.04.04 at 00:50:45 (c=0, p=9)

arXiv logo

Inference-Time Scaling for Generalist Reward Modeling

Reinforcement learning (RL) has been widely adopted in post-training for large language models (LLMs) at scale. Recently, the incentivization of reasoning capabilities in LLMs from RL indicates that $\textit{proper learning methods could enable effective inference-time scalability}$. A key challenge of RL is to obtain accurate reward signals for LLMs in various domains beyond verifiable questions or artificial rules. In this work, we investigate how to improve reward modeling (RM) with more inference compute for general queries, i.e. the $\textbf{inference-time scalability of generalist RM}$, and further, how to improve the effectiveness of performance-compute scaling with proper learning methods. For the RM approach, we adopt pointwise generative reward modeling (GRM) to enable flexibility for different input types and potential for inference-time scaling. For the learning method, we propose Self-Principled Critique Tuning (SPCT) to foster scalable reward generation behaviors in GRMs through online RL, to generate principles adaptively and critiques accurately, resulting in $\textbf{DeepSeek-GRM}$ models. Furthermore, for effective inference-time scaling, we use parallel sampling to expand compute usage, and introduce a meta RM to guide voting process for better scaling performance. Empirically, we show that SPCT significantly improves the quality and scalability of GRMs, outperforming existing methods and models in various RM benchmarks without severe biases, and could achieve better performance compared to training-time scaling. DeepSeek-GRM still meets challenges in some tasks, which we believe can be addressed by future efforts in generalist reward systems. The models will be released and open-sourced.

arxiv.org · arXiv.org

0

i think my FreeBSD bridge(4) changes are now mostly complete: github.com/freebsd/freebsd-src

summary:

- an interface's PVID can be set using 'ifconfig bridge0 ifpvid ix0 20'; any pvid other than 0 enables VLAN filtering.

- an interface's port type can be set using 'ifconfig bridge0 iftype ix0 <access|trunk|hybrid>'; the default is hybrid.

- the port's allowed VLAN list for trunk/hybrid ports can be set using ‘ifconfig bridge0 +ifvlans ix0 10,20,30,200-399', or '-ifvlans' to remove

- .1q tags are added on incoming frames and stripped on outgoing frames based on port PVID

- you can create an SVI on a bridge using 'ifconfig vlan20 create vlan 20 vlandev bridge0’ (or just 'ifconfig bridge0.20 create’)

- in theory this lets us do standards-based RSTP (rather than PVST) but i haven't tested that at all

this needs a *lot* of testing and review; i'm sure there are bugs and unhandled edge cases. my plan is to divide it into smaller chunks and submit those one at a time; so far, that's the ifconfig ifpvid option.

[edited to remove the link to a specific commit which is no longer valid]

0

ユン支持者の集会、罷免判決で怒りに燃え上がるでもなく意気消沈して解散したらしい😅
おかげで心配された混乱はなく、ケガ人も死者も出なかった。
「アメリカが守ってくれると思ったのに」と語った支持者がいたっちゅうのがなんとも。なんでトランプがそんなことすると思うのか…。
youtube.com/watch?si=8gI7-c1X-

0
0
0
0
0
0
0
0

일본에 막 살기 시작했던
지인이 너무나도 맛있다길래
궁금해서 시켜먹은 뒤로

일본에서 제일 좋아하게 된
사이제리아의 이탈리안 푸딩
:explosion_pudding:​ ​:blobcatmmm:

0
0