RandyOrion

6 ms·

31.
by RandyOrion 5mo ago
Although the performance claim of 8b dense matching 32b moe is somewhat questionable, thank you granite team for releasing small dense LLMs.
32.
by RandyOrion 6mo ago
Thank you Qwen team. Small DENSE LLMs shapes the future of local LLM users. When Qwen 3.5 27b released, I didn't really understand why linear attention is used instead of full attention because of the performance degradation and proble
33.
by RandyOrion 6mo ago
Check out Fig. 6 in this paper, it shows the comparison between the proposed method and pytorch native FSDP offload method.
34.
by RandyOrion 6mo ago
No open weights. Besides, I'm old enough to recall that META has trained a version of LLAMA 4 specifically for LM arena elo benchmaxxing and PR things, and proceeded to release a different version of LLAMA 4.
35.
by RandyOrion 6mo ago
Thank you Gemma team for releasing small dense VLM(s). The elo ranking [1] is too good to be true. I don't know why gemma-4-26b-a4b performs better than gemma-4-31b. Also waiting for more bugfixes in llama.cpp, sglang and vllm to do pr
36.
by RandyOrion 6mo ago
Please no. If you want to install APKs directly on Android phones selling in China, you'll face even more draconian restrictions imposed by both Chinese OEMs and Chinese government, e.g., cannot install telegram [1], cannot install VPN
37.
by RandyOrion 6mo ago
Yeah, let's hold Google accountable. Is there a way to practice anti-trust laws?
38.
by RandyOrion 6mo ago
Thank you for standing against the Android Developer Verification enforced by Google. Now in addition to stopping using Youtube, replacing chrome with ungoogled chromium, I'm moving to de-googled AOSP builds, e.g., lineageOS, insted of
39.
by RandyOrion 6mo ago
Wow, just wow. 1.5M records of PRs affected. Does Microsoft copilot ask users for the permission of adding ads inside their PRs before actually doing the thing? Do users show their consents on this matter? Now EVERYONE can see ads disguised
40.
by RandyOrion 7mo ago
This project shows an interesting automated search for engineering problems that I like to see more. The experience of utilizing tiered storage (gpu vram, ram, and ssd) is generally poor for a lot of LLM inference engines out there, e.g., l
41.
by RandyOrion 7mo ago
First, thank you Junyang and Qwen team for your incredible work. You deserve better. This is sad for local LLM community. First we lost wizardLM, Yi and others, then we lost Llama and others, now we lost Qwen...
42.
by RandyOrion 7mo ago
Well, I don't have enough knowledge on the boot process of RPi. However, I do expect that most modern hardware, e.g. x86, do not work like RPi, so your words do not hold in most realistic scenarios, at least for now. Besides, do curren
43.
by RandyOrion 7mo ago
As far as I know, ESNI was pushed to a large number of websites in a very fast pace, and soon completely blocked by GFW. Then I didn't hear news about ESNI.
44.
by RandyOrion 7mo ago
Cool. However, one still need CPU to send commands to GPU in order to let GPU do CPU things.
45.
by RandyOrion 7mo ago
TLS Encrypted Client Hello (ECH) standard is another attempt on encrypting plaintext Server Name Indication (SNI). It is very useful for circumventing SNI-based censorship, which is adopted for years by state-backed systems like the Great F
46.
by RandyOrion 7mo ago
Google's stance on MV2 and MV3 is definitely affecting every chromium-based projects. > but the removal of MV2 from multiple parts of the codebase is imminent. Is there any evidence showing that ungoogled chromium project is conside
47.
by RandyOrion 7mo ago
> Every single chrome-fork has shut down MV2 extensions Ungoogled chromium still supports MV2, and uBlock origin extension works fine.
48.
by RandyOrion 8mo ago
There is a term called low-trust society, and at least 1.4 billion people are struggling in that.
49.
by RandyOrion 8mo ago
A related paper interviewing victims of the pig-butchering scams. “Hello, is this Anna?": Unpacking the Lifecycle of Pig-Butchering Scams [1] [1] https://arxiv.org/abs/2503.20821
50.
by RandyOrion 8mo ago
Sadly, this already happens in China. After many years of heavy censorship on 1989 Tiananmen Square massacre, there are a lot of accounts on Chinese BBS who say all the footage "AI generated".
51.
by RandyOrion 8mo ago
NO. They can do unconditional blocking at any moment and suddenly you can experience Internet blackout. [1] The censorship from GFW is ever evolving. See the endless cat-and-mouse games yourself. [2][3] [1] https://github.com&#x2
52.
by RandyOrion 9mo ago
There is a dedicated thread of record and anti-censorship resources of the Iran Network Shutdown [1]. Hope it helps. [1] https://github.com/net4people/bbs/issues/561
53.
by RandyOrion 9mo ago
Using Lawnchair for more than a year now. No major issues on my device with Android 13 and Lineage OS 22.2. Will not look at other alternatives for a long time.
54.
by RandyOrion 9mo ago
Hi, there. Two questions about this repo [0]. Can you show us what you did after people failed to compile that project [1]? There are also questions about the attribution of these commits [2]. Can you share some information? [0] https:&#x2
55.
by RandyOrion 10mo ago
https://hn-wrapped.kadoa.com/RandyOrion Well, as a local VRAM libertarian, to manually prune the safety alignment part of a 500B LLM for it to run on 1GB RAM or VRAM is definitely a lifetime goal for me.
56.
by RandyOrion 10mo ago
The outage happened in Aug 19th-20th, China is directly linked to the misconfiguration of GFW. It's government directed, not so-called "technical problem". See https://gfw.report/blog/gfw_unconditional_rs
57.
by RandyOrion 10mo ago
I see 3-4% of 1GB ram usage for caddy only. Note that in my case caddy serves as reverse proxy and there is also very little traffic.
58.
by RandyOrion 10mo ago
Well, I use Arch Linux and the caddy package from pacman just works. You may checkout https://github.com/caddyserver/xcaddy for custom caddy build. Besides, I don't use wildcard certs. I use caddy to reverse proxy
59.
by RandyOrion 10mo ago
Thank you Let's Encrypt, together with the acme.sh , caddy and the whole ecosystem for TLS. You simply cannot emphasize the information security enough if all your Internet traffic is audited, censored and manipulated by a number of ad
60.
by RandyOrion 10mo ago
> From their project page: > We analyze over 1,100 deep neural networks—including 500 Mistral-7B LoRAs and 500 Vision Transformers. We provide the first large-scale empirical evidence that networks systematically converge to shared, l

More ›