Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
gpjt
searching Neon…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
6 ms
·
1.
▲
Extending Raschka's GPT-2: an MoE trained from scratch on an RTX 3090
(gilesthomas.com)
1 points
by
gpjt
6d ago
|
0 comments
2.
▲
Putting my Jax-trained models on the Hugging Face Hub
(gilesthomas.com)
2 points
by
gpjt
13d ago
|
0 comments
3.
▲
Why do OpenAI's GPT-2 weights beat mine? Part four: digging into dropout
(gilesthomas.com)
1 points
by
gpjt
20d ago
|
0 comments
4.
▲
First patient to undergo live AI-assisted brain surgery has tumour removed
(bbc.com)
6 points
by
gpjt
21d ago
|
0 comments
5.
▲
Adding diagrams to my static site generator with D2
(gilesthomas.com)
3 points
by
gpjt
22d ago
|
0 comments
6.
▲
by
gpjt
23d ago
In my experience, having a solid understanding of the next level down in the stack -- the foundation you're building your startup on -- is really helpful. We built a PaaS, and knowing enough about Linux internals to be able to work ou
7.
▲
by
gpjt
23d ago
But you can train a small LLM with a gaming graphics card -- I managed one on a GTX 1660. I don't think pg is suggesting that you try to chase the frontier. It's more like building your own OS in the 80s, or web server in the 90s
8.
▲
Use the built-in GELU, don't roll your own
(gilesthomas.com)
1 points
by
gpjt
28d ago
|
0 comments
9.
▲
by
gpjt
1mo ago
Hmm. Great post, not so sure about the title. I'd say (and I think this is actually compatible with the article itself, if not the title) that a better rule is to write about things that you've only just started understanding. Th
10.
▲
by
gpjt
1mo ago
I must admit that I found it confusing that they lumped people with unmanaged hypertension in with those who were on meds to manage it. Those feel like separate groups. Does anyone here know more about this area, and why the combined grou
11.
▲
by
gpjt
1mo ago
I'm adding MoE support to the GPT-2 code from Sebastian Raschka's "Build a Large Language Model (From Scratch)". Just the minimal changes. Planning to do (or at least start!) a from-scratch base model training run on m
12.
▲
A Quick(ish) Chinchilla Check
(gilesthomas.com)
1 points
by
gpjt
1mo ago
|
0 comments
13.
▲
by
gpjt
2mo ago
Lovely bit of inadvertent editorialising by the submission form there :-) Title should be (as it is on the page) " How I use AI on this blog".
14.
▲
I use AI on this blog
(gilesthomas.com)
1 points
by
gpjt
2mo ago
|
1 comments
15.
▲
Why do OpenAI's GPT-2 weights beat mine? Part three: testing overtraining
(gilesthomas.com)
2 points
by
gpjt
2mo ago
|
0 comments
16.
▲
Why do OpenAI's GPT-2 weights beat mine? Part two: the bugfix
(gilesthomas.com)
8 points
by
gpjt
2mo ago
|
0 comments
17.
▲
Why do OpenAI's GPT-2 weights beat mine?
(gilesthomas.com)
4 points
by
gpjt
2mo ago
|
0 comments
18.
▲
Benchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090
(gilesthomas.com)
16 points
by
gpjt
2mo ago
|
0 comments
19.
▲
Building intuition about LLM parameter counts
(gilesthomas.com)
2 points
by
gpjt
2mo ago
|
0 comments
20.
▲
Poppy the training box, part 1: the beginnings
(gilesthomas.com)
3 points
by
gpjt
2mo ago
|
0 comments
21.
▲
From bigrams to GPT-2, one component at a time (in Jax)
(gilesthomas.com)
1 points
by
gpjt
2mo ago
|
0 comments
22.
▲
Building a Jax training loop for an LLM training run
(gilesthomas.com)
2 points
by
gpjt
3mo ago
|
0 comments
23.
▲
Thoughts on Role Confusion
(gilesthomas.com)
3 points
by
gpjt
3mo ago
|
0 comments
24.
▲
by
gpjt
3mo ago
In the EU, at least one of the problems is regulatory/tax for digital services. They need to charge you the VAT rate for the country where you are based. For that, they need two pieces of evidence about your location. There are various
25.
▲
by
gpjt
3mo ago
OP here -- yes indeed. I ever do a new series of posts on upgrading my network to anything faster, the first one will probably be titled something like "25Gb/s Ethernet: how much it costs to rip out your CAT-6A and replace it with
26.
▲
Flax debugging: making a hash of things
(gilesthomas.com)
2 points
by
gpjt
3mo ago
|
0 comments
27.
▲
by
gpjt
3mo ago
OP here -- yes, 100%! Within my study it's DACs all the way. I only use 10GBASE-T where it's the only option: through the wall cabling, and from the connector on my ISP's crappy router to the mini-PC that acts as the real r
28.
▲
by
gpjt
3mo ago
Thanks for making the comment! It saved me having to bolt a USB fan over the old, misbehaving SFP+ module or something similarly silly...
29.
▲
10Gb/s Ethernet: switching to a Broadcom SFP+ module
(gilesthomas.com)
195 points
by
gpjt
3mo ago
|
170 comments
30.
▲
Jax: Commitment Issues
(gilesthomas.com)
4 points
by
gpjt
3mo ago
|
0 comments
More ›