ubutler

8 ms·

61.
Show HN: Terge – an easy-to-use library for merging AI models(github.com)
2 points by ubutler 2y ago|0 comments
62.
Show HN: Terge – an easy-to-use library for merging AI models(github.com)
1 points by ubutler 2y ago|0 comments
63.
Introducing the Open Arabic LLM Leaderboard – Hugging Face(huggingface.co)
1 points by ubutler 2y ago|0 comments
64.
Show HN: EmuBert – the first open encoder model for Australian law(huggingface.co)
2 points by ubutler 2y ago|0 comments
65.
by ubutler 2y ago
Although worth noting that more recent techniques (eg, Transformers) need stop words for context.
66.
by ubutler 2y ago
There are two reasons I can think of why someone might reuse a tokeniser: 1. They want to continue pretraining a model instead of starting from scratch. But actually people might not know that you can pretty easily reuse model weights even
67.
by ubutler 2y ago
I’ve trained tokenizers on medium-sized datasets (+5GB of text, although that could be considered small or large depending on who you ask) and have always found training quite fast. As in, it takes a couple minutes. Maybe if we’re talking t
68.
by ubutler 2y ago
It surprises me that this is a surprise, I thought this was a given.
69.
by ubutler 2y ago
Also, I’m not too sure that saying ‘ok’ is to ‘affirm or to express agreement’. Sometimes it can be used to simply acknowledge that someone has said something, such as when you say ‘ok sure’ or ‘ok but’. And if I am wrong, and it does mean
70.
by ubutler 2y ago
Check out PaCMAP, AFAIK it’s the current SoTA in dimensionality reduction: https://github.com/YingfanWang/PaCMAP
71.
by ubutler 2y ago
The LLM’s logits should be translatable into probabilities although I’m not too sure how meaningful those might be as models can sometimes be quite confident in entirely invalid predictions.
72.
Backpropagation from the Ground Up [video](youtube.com)
2 points by ubutler 2y ago|0 comments
73.
by ubutler 3y ago
Thanks for sharing this, it reminded me that I needed to activate DNSSEC for my domain, which I've done now.
74.
by ubutler 3y ago
This reminds me of an episode of Map Men where they offhandedly point out that by plotting British place names of Viking origins, it is possible to see the true extent of their invasion: https://youtu.be/uYNzqgU7na4?si=l-b6-
75.
by ubutler 3y ago
@dang I think this is the same story as another post published an hour earlier: https://news.ycombinator.com/item?id=39967707
76.
by ubutler 3y ago
(2018)
77.
PaperBack: How to store arbitrary data on A4 sheets of paper (2007)(ollydbg.de)
131 points by ubutler 3y ago|40 comments
78.
by ubutler 3y ago
See also xkcd #2347: https://xkcd.com/2347/
79.
How one programmer broke the internet by deleting a tiny piece of code (2016)(qz.com)
9 points by ubutler 3y ago|1 comments
80.
Over 170K users caught up in poisoned Python package ruse(theregister.com)
4 points by ubutler 3y ago|0 comments
81.
by ubutler 3y ago
The dimensionality reduction model I used, PaCMAP, had only a single parameter I could tweak, which I did not. However, it is somewhat non-deterministic AFAIK, so the final arrangement can vary on each run, but in my experience, the arrange
82.
by ubutler 3y ago
> I've read a number of acts for myself, procedural civil and criminal, and have even run into the invisible wall between legislation and case law. Glad to hear it corresponded with your lived experience, it really was surprising to
83.
by ubutler 3y ago
> Furthermore, in common law countries like Australia, you have many legal concepts that have developed independently of statute and either remain independent or have been formalised into statute. This is the bigger point. In my own univ
84.
by ubutler 3y ago
Thank you :). Would you mind sharing what you have in mind? I haven't come across a visualised semantic map of state and federal Australian laws, cases and regulations before.
85.
by ubutler 3y ago
As I note in my article, the language and style employed in Australian judgments is different from that employed in statute. Furthermore, in common law countries like Australia, you have many legal concepts that have developed independently
86.
by ubutler 3y ago
Interesting, did you try also using PaCMAP or UMAP for dimensionality reduction? It might result in a more meaningful representation of their underlying semantic structure: see the 'mammoth' example in my article.
87.
by ubutler 3y ago
Thank you :)
88.
by ubutler 3y ago
Nope, the map is built atop the Open Australian Legal Corpus, which is the first open database of Australian law (you can read about how I built it here: https://umarbutler.com/how-i-built-the-largest-open-database... ). Unf
89.
by ubutler 3y ago
I'm glad you appreciated that touch :) Seeing as 59% of my readers are on mobile, I thought it'd be better to have a static image rather than an interactive map which would be pretty unusable on a phone.
90.
by ubutler 3y ago
> It definitely provides a pretty picture, but just wanted to emphasise the map !== territory addage. You're right — my map does not necessarily represent the underlying semantic structure of Australian law, it is an approximation

More ›