peterstjohn

6 ms·

1.
by peterstjohn 3mo ago
Okay, a couple of hours later…thanks for the hint as that's fucking dark magic ;) and I now have access to the entire New Yorker again after around 15 years :)
2.
by peterstjohn 3mo ago
Ooooh, you don't happen to have the code for the New Yorker decryption in a form you could send, do you? Or put up on github or even just give me the starting prompt…
3.
by peterstjohn 1y ago
Try Hey Duggee - it's not as explicitly British-coded, but there's a ton of stuff in there if you were watching Spaced in your late teens and now find yourself a parent…
4.
by peterstjohn 1y ago
For my sins, I didn't actually realise how great that was until quite a bit afterwards! ;)
5.
by peterstjohn 1y ago
Ha, same here! It really helped my imposter syndrome, as I overheard a couple of guys talking about the ARM assembly they were doing on their Archimedes on the first day…and I hadn't written anything fancier than QuickBASIC at the time
6.
by peterstjohn 2y ago
I even hosted a mirror of the original Mozilla source code dump from St. Anselm Hall, and nobody ever complained ;P
7.
by peterstjohn 2y ago
If you think that of Owen's output, for heaven's sake I fear for you if you ever read a Jonathan Meades article…
8.
by peterstjohn 2y ago
You would just film almost _directly_ across the river and shoot on the South Bank, one of the major brutalist outposts in London. It's lovely.
9.
by peterstjohn 3y ago
I no longer work there, but Lucidworks has had embedding training as a first-class feature in Fusion since January 2020 (I know because I wrapped up adding it just as COVID became a thing). We definitely saw that even with just slightly out
10.
by peterstjohn 3y ago
Well, why wouldn't they sell (license) the rights to make Transformers films (which as far as I know is just extending their existing contract with Paramount)? They still own the underlying IP[^1], so as long as the contract is a decen
11.
by peterstjohn 3y ago
+1 to everybody that mentioned that Vespa has great vector support _and_ lexical filtering. And you likely will end up needing both. Don't sleep on some of its newer features like multi-vector document fields, either…
12.
by peterstjohn 3y ago
That paper does a terrible job of making Lucene look useful, though. 10qps from a server with 1TB of RAM is not great (and I know Lucene HNSW can perform better than that in the real world, so I am somewhat mystified that this paper is bein
13.
by peterstjohn 3y ago
It definitely depends on your use case. If you are just searching through the entire array at all times, then this is certainly an acceptable option (you could even flip it all onto a GPU too). But when you start to require filtering or com
14.
by peterstjohn 3y ago
Yes! We've been running Milvus in production for about three years now, powering some customers that do have queries at that scale. It has its foibles like all of these systems (the lack of non-int id fields in the 1.x line is maddenin
15.
by peterstjohn 3y ago
Are they forking Lucene or somehow getting the Lucene devs to increase that limit? Because this PR has been open for over a year now: https://github.com/apache/lucene/issues/11507
16.
by peterstjohn 3y ago
Fun project, with a bit of a kicker as I see the words "Colerain Avenue" and realize it was literally across the road from me.
17.
by peterstjohn 3y ago
So just use their base model and fine-tune with a non-restrictive dataset (e.g. Databricks' Dolly 2.0 instructions)? You can get a decent LoRA fine-tune done in a day or so on consumer GPU hardware, I would imagine. The point here is t
18.
by peterstjohn 3y ago
I once travelled with a 5kg vat of fondant icing on a transatlantic flight. "Yes, it looks very much like Semtex, but it's fine!" Still not exactly sure how I got away with it…
19.
by peterstjohn 4y ago
Heh, my eyes did pop at that one, considering we've also been doing that over here since 2020 at least ;)
20.
by peterstjohn 4y ago
It really does give you the best of both worlds - resistant to typos, handling synonyms without all the usual hand-written rules, but still able to handle direct searches like ISBNs. (disclaimer: I work on Semantic Search at Lucidworks)
21.
by peterstjohn 4y ago
Two big reasons for Vespa over Milvus 1.x: * Filtering * String-based IDs (a caveat that I haven't used Milvus 2.x recently, which does fix these issues, but brings in a bunch of other dependencies like Kafka or Pulsar)
22.
by peterstjohn 4y ago
UTAH SAINTS! UTAH SAINTS! ;P (It would have been more fun if we'd spent the past month with clickbait like "What is Orgone Energy, Anyway?"
23.
by peterstjohn 5y ago
If you control the HNSW implementation, it can definitely do pre-filtering. Vespa does it, and you can modify open source HNSW libs easily. I added pre-filtering support to an internal fork of HNSWLIB last week, for example…
24.
by peterstjohn 5y ago
It was the Gameboy version, not the C64. https://www.youtube.com/watch?v=wGIKnn-COS4
25.
by peterstjohn 5y ago
It's not a lie, yes, but it does rather undermine the entire point of the machine if you have to re-temper outside of it to use a standard shaped mold.
26.
by peterstjohn 5y ago
The ring molds look quite awkward too…
27.
by peterstjohn 5y ago
Two hours is _really_ fast (normally you end up grinding for 8+ hours), so I'm curious as to what the ball mill does differently than other melangeurs. The yield is quite low (250g) when you consider you can easily get 2kg out of a Pre
28.
by peterstjohn 6y ago
There are a few recycled interviews from The Living Dead in ep 2 for sure.
29.
by peterstjohn 8y ago
We both know Durham is better though ;P
30.
by peterstjohn 8y ago
Any evidence that Ferrero is planning to drop Kinder Surprise?

More ›