t55

6 ms·

91.
The Desire to Be Liked Is Rotting Your Brain(flocrivello.com)
2 points by t55 2y ago|0 comments
92.
Thou Shalt Not Overfit(argmin.net)
1 points by t55 2y ago|0 comments
93.
Obsidian's Web viewer lets you open external links within Obsidian(obsidian.md)
2 points by t55 2y ago|0 comments
94.
Quaternions and spherical trigonometry(terrytao.wordpress.com)
124 points by t55 2y ago|48 comments
95.
Qwen2.5-VL: State-of-the-art multimodal LLM(github.com)
3 points by t55 2y ago|0 comments
96.
Goose – an open-source, extensible AI agent that goes beyond code suggestions(github.com)
40 points by t55 2y ago|5 comments
97.
I don't believe DeepSeek crashed Nvidia's stock(understandingai.org)
3 points by t55 2y ago|1 comments
98.
by t55 2y ago
Totally agree!
99.
Large Language Model Training Using FP4 Quantization(arxiv.org)
2 points by t55 2y ago|0 comments
100.
Supervised Fine-Tuning Memorizes, RL Generalizes(arxiv.org)
1 points by t55 2y ago|0 comments
101.
by t55 2y ago
Thank you so much, this is super valuable feedback and actually very easy to add. We will add this soon!
102.
by t55 2y ago
TFv1 or TFv2? AFAIK it's batch-first in TFv2
103.
by t55 2y ago
Same here lol! I ask it to write in bullet points every time
104.
by t55 2y ago
Ah, good catch! You seem to really assume the worst on our side, but fine, this is HN :) The author was simply there because of the website template we used; by default, it wants you to specify an author so we did. I removed the author now,
105.
by t55 2y ago
Would be curious to compare the results!
106.
by t55 2y ago
Hi, OP here; this article helped me a lot to understand better KV caches, which is ultimately why I co-wrote it with AI + read it several times before posting
107.
by t55 2y ago
I did :) Yes, 80% of the post is generated, yet, I still reviewed everything! And if I had written everything from scratch, this would have probably taken a week rather than a few hours.
108.
by t55 2y ago
The phrase, “the first token looks at 1 token,” is simply a shorthand for the self-attention step when the sequence length is one. Although there are no preceding tokens, we still treat it as an O(1^2) operation where the first token effect
109.
by t55 2y ago
> the contents are pretty good, they cover many papers in a lightweight way, you can get a decent overview in 10 minutes for what would take hours to research. Exactly, I'm still surprised it works so well. Also, which formatting do
110.
by t55 2y ago
Yep, me too. Always loved to write in bullet points and in fact, I prompted the LLMs explicitly to do so.
111.
by t55 2y ago
Hi, OP here. My intention wasn't to "gotcha" anyone by mentioning that in the end, it was simply to be upfront. Many blog posts/content put out these days are obviously 100% AI-generated, yet it's never being mentio
112.
by t55 2y ago
I see! We wanted an overview of KV caching internally and quickly understand the latest methods without much fluff, I think it did a great job at that
113.
by t55 2y ago
We will push it to our repo very soon :)
114.
by t55 2y ago
It’s mostly a convention. In many deep learning frameworks (PyTorch, TensorFlow, etc.), inputs are stored with the “batch × length × hidden-dim” shape, effectively making the token embeddings row vectors. Multiplying “xW” is then the natura
115.
by t55 2y ago
let me know if you have more feedback!
116.
by t55 2y ago
Replaced the exclamation point with a dot, hope it's better now!
117.
by t55 2y ago
We do support loops! :) Just haven't advertised it inside the Github readme yet. One of the reasons why we started building pyspur is precisely because none of the other tools support loops! If you need more support; shoot me an email;
118.
by t55 2y ago
Thanks! Yes, will push it as template to our repo ( https://github.com/PySpur-Dev/pyspur ) soon! We used o1 and claude3.5
119.
by t55 2y ago
It doesn't matter whether you use AI or not, what matters is whether your articles are relevant and useful or not
120.
by t55 2y ago
Just refined it a bit; I hope it's clearer now!

More ›