jxy

11 ms·

91.
by jxy 3y ago
It's great that Glibc documents its POSIX violating behavior very well, but it doesn't change the fact that it violates POSIX.
92.
by jxy 3y ago
Either Linux's fault or Glibc's fault. POSIX mandates getaddrinfo to be thread safe.
93.
by jxy 3y ago
ehhh, the design of the webpage infuriates me. Who thought sprinkling faint dots swimming across the webpage was a good thing? I almost thought something's wrong with my eyes or my display!
94.
by jxy 3y ago
Is there an option to use the API and just pay per use? I couldn't find it.
95.
by jxy 3y ago
It's kind of true with git-lfs.
96.
by jxy 3y ago
Does the hardware requirements make sense to you? > Training QLoRA 160GB 2x A100 80GB > Inference GPTQ/int4 320GB 8x A100 40GB Why would inference in int4 require more memory than training with QLoRA, which I believe is also 4bit
97.
by jxy 3y ago
Right. ### off topic rants below Somehow there are so many blogpost about these things, all trying to ask for your emails. Is it becoming easier to put more words together nowadays? I guess so. I really wish there is a way to fact check all
98.
by jxy 3y ago
That's the original William Strunk Jr.'s version. The Strunk & White version is not in public domain yet.
99.
by jxy 3y ago
Their code README says > PreTeXt is guided by the following principles: > 1. PreTeXt is a markup language that captures the structure of textbooks and research papers. > 2. PreTeXt is human-readable and human-writable. When did XML
100.
by jxy 3y ago
Would it allow CUDA support?
101.
by jxy 3y ago
Would this be ported to Safari?
102.
by jxy 3y ago
What data leaves my iPhone?
103.
by jxy 3y ago
The Photos app already automatically OCR everything and you can search through any images including screenshots.
104.
by jxy 3y ago
I want something like this but for the whole display. Due to the security implications, obviously it would need the OS native support.
105.
by jxy 3y ago
Does it run on any of the BSDs?
106.
by jxy 3y ago
Where's CUDA on FreeBSD, Nvidia?!
107.
by jxy 3y ago
Thanks for the link. The decision of burying these extra information in a support article, not cool!
108.
by jxy 3y ago
You can run whisper.cpp locally in real time: https://github.com/ggerganov/whisper.cpp/tree/master/example...
109.
by jxy 3y ago
They didn't mention gpt-4-32k. Does anybody know if it will be generally available in the same timeframe? There's still no news about the multi-modal gpt-4. I guess the image input is just too expensive to run or it's actual
110.
by jxy 3y ago
They said in the post, > We recognize this is a significant change for developers using those older models. Winding down these models is not a decision we are making lightly. We will cover the financial cost of users re-embedding content
111.
by jxy 3y ago
> The MI250 has a larger amount of HBM memory (128GB) than even the largest A100 (80GB). This means the MI250 can hold larger models for training or inference. AMD put two individual devices together, so you can tell people the combined
112.
by jxy 3y ago
It seems to use a different tokenizer than LLaMA, though the neural network architecture is the same.
113.
by jxy 3y ago
The walk(1) on 9front seems to have grown up and supports many options: http://man.9front.org/1/walk
114.
by jxy 3y ago
The new kid just learned to walk(1). http://git.9front.org/plan9front/plan9front/HEAD/sys/src/cmd... 329 lines of code.
115.
by jxy 3y ago
They are clearly scared of ggml.
116.
by jxy 3y ago
1 to 2 tokens per second on CPU, depending on the CPU/RAM, thanks to GGML. See discussions: https://github.com/ggerganov/ggml/pull/231 And the code: https://github.com/jploski/ggml&#
117.
by jxy 3y ago
The fundamental theories we have is Quantum Electrodynamics and Quantum Chromodynamics. This new experiment tells you that "modern nuclear forces, including those derived within chiral effective field theory" break down and cannot
118.
by jxy 3y ago
Over 300 million reside in a nation with easy access to guns, leading to corporate profits and fatalities, while legislators focus on leaked company figures.
119.
by jxy 3y ago
Any POSIX system that has a C++11 compiler could compile llama.cpp and run it, including all the BSDs. The OS supported by onnxruntime is very limited.
120.
by jxy 3y ago
I'm waiting for The Amendment: A well regulated AI, being necessary to the security of a free State, the right of the people to keep and bear GPUs, shall not be infringed.

More ›