pulse7

7 ms·

1.
by pulse7 6d ago
Prompt processing on Macs is VERY slow...
2.
by pulse7 7d ago
I hope we will soon have an open-source project for training such small LLMs where one can just pick the architecture (like Qwen / DeepSeek / etc.), parameter count, dataset, ... and then let it run on a local/rented GPUs...
3.
by pulse7 21d ago
"code translation in hardware" microcode is called "native"
4.
by pulse7 1mo ago
So it's mostly the "Optionality". Like USB. And yet USB is everywhere...
5.
by pulse7 1mo ago
It there anything similar for RTX 3090 and RTX 4090?
6.
by pulse7 1mo ago
Can you please tell which Gemma 4 variant managed to correctly reason through your private benchmarks? Was is Gemma 4 31B? What quantizations and context lengths did you use for Gemma 4 and Qwen 3.8 27B? I am asking because I can't eve
7.
by pulse7 1mo ago
"--context-shift, --no-context-shift ... whether to use context shift on infinite text generation (default: disabled)" From: https://github.com/ggml-org/llama.cpp/blob/master/tools/serv...
8.
by pulse7 1mo ago
It will come... all big hardware players (Intel, AMD, Broadcom) and dozens of startups (Tenstorrent, etc.) are working on it...
9.
by pulse7 1mo ago
Hasn't Docker always been just a thin layer of duct tape over existing solutions?
10.
by pulse7 1mo ago
You cannot wear out an SSD through AI inference alone. LLM weights are only read from the SSD, and read operations do not contribute to SSD wear. SSD wear is primarily caused by write and erase operations.
11.
by pulse7 2mo ago
Those are routers - not general purpose computers.
12.
by pulse7 2mo ago
On the absolute limit of ethics? This is simply not ethical anymore...
13.
by pulse7 2mo ago
The same here! Merged several repos into monorepo are retained full history!
14.
by pulse7 2mo ago
Most probably not optimized yet for this model...
15.
by pulse7 2mo ago
With electron-beam lithography you can build transistors with gate lengths down to 1 to 3 nanometers.
16.
by pulse7 3mo ago
exactly: high demand!
17.
by pulse7 3mo ago
NVidia Spark is much slower (low memory bandwidth)!
18.
by pulse7 3mo ago
Can you please share you llama.cpp server parameters to turn on modern LLM sampling stack? Docs [1] say that the top_n_sigma is already in the default sampler list: "(default: penalties;dry;top_n_sigma;top_k;typ_p;top_p;min_p;xtc;tempe
19.
by pulse7 3mo ago
This https://en.wikipedia.org/wiki/Elon_Musk%27s_Tesla_Roadster ?
20.
by pulse7 3mo ago
I'm sorry your dad didn't respect your IT work...
21.
by pulse7 3mo ago
So it's similar to "Andy and Bill's Law" [1]: "What Intel giveth, Microsoft taketh away". If Windows would stay the same (and not grow) it would be much faster on newer CPUs... [1] https://en.wikiped
22.
by pulse7 5mo ago
It looks like the president - which was a businessman - will make a huge damage to American IT businesses. And IT stocks dominate the S&P 500, comprising roughly 1/3 of the index's total market capitalization... Good luck Amer
23.
by pulse7 6mo ago
Source?
24.
by pulse7 7mo ago
Maybe they can stack LLM parameters in 200 layers like 3D NAND flash and make the chip very small ...
25.
by pulse7 10mo ago
IBM was founded in 1911 and it survived many things...
26.
by pulse7 10mo ago
...and there would be dozen equally capable open-weight models which could be run locally at almost no cost... poor AI investors in this case..
27.
by pulse7 10mo ago
Nvidia is the ultimate beneficiary of the money invested (due to expensive GPUs). If Nvidia loses these good customers, it will have less revenue. So it prefers to slowly buy it's customers with this money...
28.
by pulse7 10mo ago
<joke> GGUF when? </joke>
29.
by pulse7 10mo ago
I'm very happy that "AGI office workers" will use Microsoft products - so I don't have to do it anymore... But: they will not pay a dime for the licenses...
30.
by pulse7 1y ago
You can put camera in transparent Faraday cage. With camera and gyro one can do basic positioning...

More ›