pama

29 ms·

541.
by pama 3y ago
Yes, GDS will accelerate the IO to the GPU. I’d love to see the above C code compared to hyperoptimized GPU code on the right hardware, but I don’t want to accidentally nerd snipe myself :-) The unfortunate part of this particular benchmar
542.
by pama 3y ago
There are some good ideas for this type of problem here: https://github.com/dannyvankooten/1brc After you deal with parsing and hashes, basically you are IO limited, so mmap helps. The C code takes less than 1.4s witho
543.
by pama 3y ago
The authors explicitly generated new random data to address this concern.
544.
by pama 3y ago
Not sure what you are looking for, but perhaps these thoughts can help. Better geometry understanding helps with spatial tasks. Better probability theory, discrete math, and linear algebra background helps with algo development. More broad
545.
by pama 3y ago
Per the text: x is a sample (see the tilde symbol) from the density q0; the Jacobian comes from the change of variables in the density. The typesetting could use some work but the definitions are in the snippet you pasted. I agree that peo
546.
by pama 3y ago
As per Steven Heidel’s tweet this version will be released to chatGPT soon.
547.
by pama 3y ago
Math is not just arithmetic and yes better math does help at least some GPT-4 users.
548.
by pama 3y ago
Agreed that detailed documentation is required anyways. However, LLMs give you the ability to measure the quality of the documentation and the improvements in quality for changes in the documentation without having to wait for complaints by
549.
by pama 3y ago
Not sure it is worth spending any effort on this subject, however a ton of existing laws in almost any juristiction could be applied to this particular crime. I doubt the perpetrators would be caught, but I also doubt the punishment would b
550.
by pama 3y ago
Thanks. Absolutely no offense taken. No worries. I am used to frank communication and it occasionally doesn’t come out well in quick writing on the phone. The main concern I have with frameworks is the inversion of control of the code logic
551.
by pama 3y ago
Thanks for these efforts. I am surprised that you “… were surprised how many built bespoke state management layers and used printlines to debug.” Print is the common debug tool since the start of CS, and bespoke state management layers, tho
552.
by pama 3y ago
Scaling in this broader context can have many meanings. All of them disagree with the original statement and with yours. If you assume that OpenAI spent more on GPT3.5 than on GPT3, and more on GPT4 than on GPT3.5, which is certainly plausi
553.
by pama 3y ago
I have seen no evidence that LLMs don’t scale. What evidence did you have in mind? Do you mean something specific and perhaps different from what I would call scaling? It seems in the last two years people got better at squeezing out more p
554.
by pama 3y ago
You are talking about the greatest exposed hack of the computer supply chain so far by a big margin. Laws can be made retroactively for this type of thing. It has implications that are beyond the legal system, as the threat level is way be
555.
by pama 3y ago
The core source of the vulnerability (symbol lookup order allowing a dependency to preempt a function) might theoretically be fixed at the Linux+OpenSSH level.
556.
by pama 3y ago
Super nice story on the matmul optimization that gave 810 gflops for 512x512. Thanks for the write up and the contributions to llama.cpp and the community more broadly.
557.
by pama 3y ago
Only correlation can never explain something. A bomb will tear a building into bricks, but finding bricks every time after a building explodes does not help identify the cause. Suppose it ended up eventually that the causal source was a n
558.
by pama 3y ago
Since the LLM sometimes generates invalid COBOL a simple practical solution would be to use an API and allow it to test its code with GnuCOBOL, feed the output and have it try again a couple of times. I wonder what the updated benchmarks w
559.
by pama 3y ago
Not sure what you mean by wrong. I have never encountered a case yet when training an LLM (no matter what architecture) would require limited memory and was pointing out that the typical memory requirements for training are much higher yet
560.
by pama 3y ago
I’ve wondered about the same point for a long time. In smaller language models of chemistry the tree search in decoding has made a huge difference [1]. If the training data set doesn’t include the right conditioning it may be hard for the L
561.
by pama 3y ago
The big (huge?) memory requirement is during training. These LLMs work with high dimensional vectors and they calculate gradients with respect to high dimensional vectors and they do updates that require state of the optimizer. If you hav
562.
by pama 3y ago
Some additional info here: https://x.com/yampeleg/status/1773364830423937090?s=46
563.
Introducing Jamba: AI21's Groundbreaking SSM-Transformer Model(ai21.com)
4 points by pama 3y ago|1 comments
564.
by pama 3y ago
Not sure what you mean? If the recipe is not right it is no major problem but so far all was great. Their trips and questions have been on the spot and it has helped them dramatically.
565.
by pama 3y ago
We’ve purchased an annual subscription for my 85 yo (long time retired) inlaws. They use it daily. They ask their doctors insightful questions and get better care. They planned trips with it. They get recipe advice. I don’t know how much
566.
by pama 3y ago
I agree that the quality of the posts has decayed dramatically since the start of the COVID pandemic and lots of the Reddit type memes and upvotes have added horrible levels of noise. I can still find gems in it and it’s still miles ahead o
567.
by pama 3y ago
There is always a first entity to start any market and always a last one to exit it. So technically it happens thousand of times per day. Surely it’s not illegal. I could decide to sell a completely useless crocodile-alike decorative robot
568.
by pama 3y ago
So if everyone else decides to exit a market, your sole product becomes illegal?
569.
by pama 3y ago
Per the Nature paper, this fat is believed to be part of an immune system response and not related to your intake or total body fat. It is also far from clear that eliminating the production or use of this fat during a microbe invasion woul
570.
by pama 3y ago
As per the Nature paper, this is an immune system mechanism with these particular “fat” particles presumed to help fend off an invading microbe. So no, it doesn’t matter how much fat or not you have to begin with as you cannot as easily red

More ›