gliptic

10 ms·

91.
by gliptic 3y ago
When style guides recommend you to use "a historic" in formal writing, why should I not talk factually about what they recommend. You were the first to complain about the usage in this article.
92.
by gliptic 3y ago
It's not. In some dialects H would be silent, and you can certainly write "an historic" because of that in informal writing, but it's discouraged in style guides that sites like this would use.
93.
by gliptic 3y ago
No, "historic" doesn't start with a vowel sound.
94.
by gliptic 3y ago
It's not NASA, but an independent group. There are others, like Everyday Astronaut.
95.
by gliptic 3y ago
> We show that the equivalence of QC techniques (with IBM, Google and others compared with our version of QC) has been hidden for about 2,500 years – since Pythagoras. Can it be any more of an obvious scam and/or ramblings of a cryp
96.
by gliptic 3y ago
Not really. You can have default values for positional arguments just fine in C#.
97.
by gliptic 3y ago
I see. I wasn't familiar with that term in this context, thanks.
98.
by gliptic 3y ago
The unstated assumption is that you also have FALSE.
99.
by gliptic 3y ago
The sign bit is the only bit involved here. What "subtractive bits" are you referring to?
100.
by gliptic 3y ago
Why would truth preservation prevent it from being functionally complete? How can you even tell a truth table is truth preserving? A truth table is not a logical argument.
101.
by gliptic 3y ago
I told you a while ago that llama.cpp supports 4/8-bit (and other) quantization just fine (and fast) on x86_64.
102.
by gliptic 3y ago
It's not a "speed limit of 0.1c for matter". This limit would apply to the scenario they investigated -- colliding black holes where the spacetime is massively affected by the mass and movement of the black holes.
103.
by gliptic 3y ago
It doesn't support Falcon right now, but there's a fork that does ( https://github.com/cmp-nct/ggllm.cpp/ ).
104.
by gliptic 3y ago
Sorry, but the topic of this post, llama.cpp, runs quantized 4/8 bit models just fine on x86_64 with AVX2, or am I missing some requirement you have?
105.
by gliptic 3y ago
The post that worries me most about the future is "Ask HN: What Happened to Pizza?"
106.
by gliptic 3y ago
It does support that. I'm using it like that on two machines.
107.
by gliptic 3y ago
They already had their own format before that.
108.
by gliptic 3y ago
No, at least C, C++, D, Rust and Zig allow static linking of the runtime libs, even on Windows. And they don't recompile it every time. Rust especially will static link everything by default.
109.
by gliptic 3y ago
There's some "artistic license" that you could also call scientific license, because you want to pick colors that maximize contrast and highlights features you want to highlight, like different atomic elements.
110.
by gliptic 3y ago
The parse function takes a string in the format mentioned.
111.
by gliptic 3y ago
No, this doesn't create previously unknown mass, and if it did it would be a dark matter _candidate_, not alternative.
112.
by gliptic 3y ago
> - Memory use aside, forward pass only is potentially a big training speed increase. Forward-pass only doesn't mean it's faster. It converges much slower than even SGD. It is a memory-time tradeoff, but that's it.
113.
by gliptic 3y ago
I cannot find any such numbers in the paper. What the paper says is that MeZO converges much slower than SGD, and each step needs two forward passes. "As a limitation, MeZO takes many steps in order to achieve strong performance."
114.
by gliptic 3y ago
> For CPU it is logical to expect 12x training speed improvement, but it is still too slow to be consistent practically useful. I don't see what you base this on. MeZO trades one back-propagation pass for another forward pass. Why w
115.
by gliptic 3y ago
Only the first link refers to this, the others are the previous "hat" tile.
116.
by gliptic 3y ago
JSON allows you to store arbitrarily large integers/floats. It's only in JS this is a problem, not if you use JSON in languages that support larger (than 54-bit) integers.
117.
by gliptic 3y ago
Somehow the most annoying thing to me is this long-standing bug where "Exit" doesn't mean "Exit", but "Exit, unless there's updates, then restart".
118.
by gliptic 3y ago
> To achieve high-precision measurements from a balloon-borne environment, the SuperBIT gondola – at roughly 3500 lbs – stabilizes its telescope to sub-arcsecond precision (akin to a three degree-of-freedom Steadicam) while sophisticated
119.
by gliptic 3y ago
Very good contrast. Usually no need for collimation. No diffraction spikes (well, SCT don't have that either).
120.
by gliptic 3y ago
Carbon fiber would be a no-brainer for weight and focus stability. It is not even that expensive.

More ›