gliptic

11 ms·

121.
by gliptic 3y ago
Depends on the lens. Lenses made for general photography often don't stack up against a good triplet astrograph when it comes to things like chromatic aberration, field flatness, sharpness etc. You will also need a good tracking mount
122.
by gliptic 3y ago
> To be semantically driven, it would be reasonable to expect that synonyms would have similar representations. How could a tokenizer do anything about that unless the synonyms actually share substrings? The vector embedding is learned,
123.
by gliptic 3y ago
What's with this widespread confusion between kW and kWh, especially from supposed experts in solar power...
124.
by gliptic 4y ago
The size mentioned is already quantized (and to integers, not floats). mmap obviously doesn't do any quantization.
125.
by gliptic 4y ago
GPTQ-for-LLaMa recently implemented some quantization tricks suggested by the GPTQ authors that improved 7B especially. Maybe llama.cpp hasn't been evaluated with those in place?
126.
by gliptic 4y ago
> We know a bunch of the people whose names are on the letter didn't sign it Yann LeCun's name never was on the letter. Where did this meme come from.
127.
by gliptic 4y ago
RWKV training can be parallelized just like transformers. It's not LSTM.
128.
by gliptic 4y ago
Correct me if I'm wrong, but I think you still need to compute gradients of non-trained weights in order to compute the gradients of the LoRA weights. What you don't have to do is store and update the optimizer state for all those
129.
by gliptic 4y ago
Lee Smolin suggested https://en.wikipedia.org/wiki/Cosmological_natural_selection But that specific hypothesis doesn't seem to match our universe anyway.
130.
by gliptic 4y ago
The page says image uploading will come later.
131.
by gliptic 4y ago
Similarily, it should not be too hard to implement checkmate. If, on your move, any of the opponent pieces can capture your king and this remains true wherever the king moves, you're mated. Should only require overlap detection between
132.
by gliptic 4y ago
This looks a lot like a functional version of Icon https://en.wikipedia.org/wiki/Icon_(programming_language)
133.
by gliptic 4y ago
In actual code it's not much difference, e.g.: while (test("-")) { element() } vs. do { element(); } while (test("-"))
134.
by gliptic 4y ago
His lawsuit against chess.com is certainly not about (2).
135.
by gliptic 4y ago
If the piece movement rules are randomized, it's not going to be learning chess, is it...
136.
by gliptic 4y ago
Why would you think it doesn't run on Linux, or that it's in C#? EDIT: Ok, it used to be a .NET application (not any more), but .NET runs fine on Linux.
137.
by gliptic 4y ago
Check out http://mattmahoney.net/dc/text.html . The top BWT performer is #22. What it comes down to is the way the BWT output is transformed and entropy-coded afterwards, and various other tricks. bzip2 only does very
138.
by gliptic 4y ago
Yes, it's not replacing 64-bit math with 80-bit math, x87 just doesn't have proper 64-bit floats with correctly sized exponent-field and the flags here interpret float literals as "long double" if I remember correctly. I
139.
by gliptic 4y ago
I'm not aware of any time GCC does that except due to C++'s promotion rules (e.g. float + double -> double), which is a problem with C++, not the compiler. I write deterministic software using floating point in C++.
140.
by gliptic 4y ago
Except the compilers will not be dumb enough to use it implicitly when you don't ask for it, just like they don't automatically use FMA now (except maybe with -ffast-math).
141.
by gliptic 4y ago
That is because of other weaknesses of bzip2, not because BWT is worse. BWT generally beats LZ-based algorithms. I would also dispute that LZMA is simpler, at least in compression. Good LZ parsing is not easy.
142.
by gliptic 4y ago
It is using the UNet, though.
143.
by gliptic 4y ago
It's specialized to &[u8] in order to supply optimized implementations.
144.
by gliptic 4y ago
It's a play on the "draw the rest of the fucking owl" meme.
145.
by gliptic 4y ago
Depends. The default script uses a non-deterministic stochastic encode for img2img. That would need to be modified as far as I know.
146.
by gliptic 4y ago
If we truly didn't know how computers came about, it's certainly progress to figure out the silicon can come from sand. How is it not origins of life research to attack every step in the process?
147.
by gliptic 4y ago
No, that is exactly what they do in opening preparation.
148.
by gliptic 4y ago
Resolution would not improve, but contrast would and you would be able to see fainter objects with less noise.
149.
by gliptic 4y ago
Yeap, this is very likely proper motion. That star is in our galaxy.
150.
by gliptic 4y ago
> HST might have done that if its instruments had been of different pixel size or imaging array size / focal length HST is already imaging at diffraction-limited resolution (with proper post-processing). It would need a bigger mirro

More ›