albertzeyer

6 ms·

31.
by albertzeyer 1y ago
This is Windows only. I wonder, why don't they use Lazarus ( https://www.lazarus-ide.org/ )? That would also make it cross-platform, and probably gain much more interest in the project.
32.
by albertzeyer 1y ago
Oh, I just checked, there is an active fork: https://github.com/ianBBB/boa-constructor
33.
by albertzeyer 1y ago
Ok, this post is mostly about text-based IDEs, but I think the point mostly stands as well for IDEs in general. I'm thinking about Visual Basic or Delphi. I think such a IDE for Python would really be helpful for beginners. Not text-ba
34.
by albertzeyer 1y ago
ICLR 2026 reviews are happening now (or soon). This paper here was accepted at ICLR 2025.
35.
by albertzeyer 1y ago
Why not store the data directly as Arrow files, to allow for mmaping? I see F3 also supports such zero-copy mmap, and skimming through the paper, it actually seems that it uses Arrow buffers, so I wonder what is the difference to directly u
36.
by albertzeyer 1y ago
Long time ago, that was actually very easy on Mac, via SIMBL ( https://github.com/albertz/simbl ) and Afloat ( https://github.com/rwu823/afloat ) and you could hack around using FScriptAnywhereSIMBL (
37.
by albertzeyer 1y ago
This sounds interesting. I would really like to read a full research paper made out of this, which describes the method in more detail, gives some more examples, does more analysis on it, etc. Btw, this uses LLMs on pure text-level? Why not
38.
by albertzeyer 1y ago
Why not use getaddrinfo_a / getaddrinfo_async_start / GetAddrInfoExW? Or just use some standalone DNS resolve code or library (which basically replicates getaddrinfo but supports this in an async way)? See also here the discussion
39.
by albertzeyer 1y ago
I keep my /etc under Git. When the system does changes automatically (via an update or whatever), I make a Git commit with a special distinct message, and so I can easily filter out all my own changes.
40.
by albertzeyer 1y ago
It's funny: The article talks about GPT and diffusion in the context of technological innovations, and then getting to AI/LLMs. But "diffusion" and "GPT" have some very different meaning in the context of AI&#x
41.
by albertzeyer 1y ago
I don't really need the versioning aspect too much, but sometimes I modify the photos a bit (e.g. rotating or so). But all the other things are relevant for me, like having it distributed, syncing, only partially having the data on a p
42.
by albertzeyer 1y ago
How much data do you have? I'm using git-annex on my photos, and that are around 100k-1M files, several TB of data, on a ZFS. In the beginning, everything was fine, but it starts to become increasingly slow, such that every operation t
43.
by albertzeyer 1y ago
I have seen such variants as well, where you actually need to install some antennas (RTK antennas?). But that was already too much effort for me, so that's why I chose the eufy. Also, from reports that I have read, it doesn't nece
44.
by albertzeyer 1y ago
Maybe 300 m^2 or so. The eufy E15 is for up to 800 m^2. There is the eufy E18 for up to 1200 m^2. I have seen other (more expensive, bigger) robots for much larger lawns. My garden is relatively flat with a few bumps here and there. You can
45.
by albertzeyer 1y ago
In my case, my lawn isn't easily accessible and also not visible from the street (because the house and garage is between the street and the garden), and I trust my neighbors. So, I think (I hope) this isn't so much an issue for m
46.
by albertzeyer 1y ago
> The current generation of robotic lawn mowers sucks. Basically all of these bots drive in a random direction until they hit the border of the lawn, rotate for a randomized duration and repeat. I recently (a few weeks ago) bought one. W
47.
by albertzeyer 1y ago
I thought that is what you mean when you said "a lot of interesting bits about the history of the development of neural networks (including backpropagation) can be found in the book Talking Nets", that there is some relevant refer
48.
by albertzeyer 1y ago
What do you mean? This popular paper is cited: [RUM] DE Rumelhart, GE Hinton, RJ Williams (1985). Learning Internal Representations by Error Propagation.
49.
by albertzeyer 1y ago
How do you not have the citations in front of you? They are all in the article? I don't expect that any relevant (re)invention of backprop is missing there. Or, if you really know some reinvention of backprop that is not mentioned here
50.
by albertzeyer 1y ago
It's not about the probability of individual tokens. It's about the probability of the whole sequence of tokens, the whole answer. If the model is good (or the human comedian is good), a good funny joke would have a higher probabi
51.
by albertzeyer 1y ago
No, it was not difficult at all. I really wonder why they have such a bad example here for GPT1. See for example this popular blog post: https://karpathy.github.io/2015/05/21/rnn-effectiveness/ That was
52.
by albertzeyer 1y ago
> Once someone hits AGI/SGI I don't think there will be such a unique event. There is no clear boundary. This is a continuous process. Modells get slightly better than before. Also, another dimension is the inference cost to ru
53.
by albertzeyer 1y ago
This is exciting. So this is using unified memory of CUDA? I wonder how well that works. Is the behavior of the unified memory in CUDA actually the same as for Apple silicon? For Apple silicon, as I understand, the memory is anyway shared b
54.
by albertzeyer 1y ago
It's a bit weird that such a paper (standard ML topic, would fit to NeurIPS, ICLR, ICML or so) is published in such a conference (IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)), which seem
55.
by albertzeyer 1y ago
The API can be used both via your normal Google account, or via API key? Because it says in the README: > Authenticate: When prompted, sign in with your personal Google account. This will grant you up to 60 model requests per minute and
56.
Show HN: Py_better_exchook: intelligently print variables in stack traces(github.com)
2 points by albertzeyer 1y ago|0 comments
57.
by albertzeyer 1y ago
"hundreds of thousands to potentially millions of tokens" - that's the same order as current commercial LLMs. Also note, if the sequence length is not really much larger than the model dimension (at least two orders of magnit
58.
by albertzeyer 1y ago
I'm not sure what type of model Google uses nowadays for their webinterface. I know that they also actually provide LLM-based translation via their API. Also the traditional cross-attention-based encoder-decoder translation models supp
59.
by albertzeyer 1y ago
Do you have any comparisons in terms of WER? I doubt that GPT-4o-transcribe is better than the best models from that leaderboard ( https://huggingface.co/spaces/hf-audio/open_asr_leaderboard ). A quick search on thi
60.
by albertzeyer 1y ago
That's not quite true. State of the art both in speech recognition and translation is still a dedicated model only for this task alone. Although the gap is getting smaller and smaller, and it also heavily depends on who invests how muc

More ›