imjonse

24 ms·

451.
by imjonse 3y ago
The second example says "First allocate 1GiB, free it, then spawn a process which does the same but with 2GiB." In the code it seems it is the parent process that does the second call too. if (pid != 0) { malloc_and_fill(1
452.
by imjonse 3y ago
I haven't heard the term before either, but it's mentioned at least by Google too so not entirely unheard of in the multimodal training context. Even so I also think this could have been conveyed without using that term. Two tower
453.
by imjonse 3y ago
My understanding is the LeCun uses the term joint embedding as an alternate name for siamese networks. It does not necessarily imply multimodality and he has used it mostly in the context of SSL for images.
454.
by imjonse 3y ago
but they say this is old tech, with exprired patents, with some confusing 'Pfizer offered to buy it' and no extra details. To me it seems they mostly want VC money from those who would bet on something unproven if the potential ma
455.
by imjonse 3y ago
If this is such a revolutionary technology and it is cheap to mass produce, it should be like water fluoridation and similar state funded measures that increase overall population health. I am not arguing that $20k is more than regular dent
456.
by imjonse 3y ago
They have a pitch deck lamenting 90% of US adults have cavities, and they are 2x likely to get untreated if you are poor, then propose to market this for 20k$/dose. Saving the world, again, startup style.
457.
by imjonse 3y ago
Since most of the hyperlinked web is not document-heavy today, as opposed to what it was 30 years ago, form followed function and we have distracting visuals and annoying UX over mostly shallow content :)
458.
by imjonse 3y ago
As you know, energy efficiency is not related to research and training only, but increasingly to inference and productionizing of the models. This is mostly what this course emphasizes too. It's not about luddites putting on the enviro
459.
by imjonse 3y ago
Do you know if there are any devices where the high FPS mode can be used not only for capture but also for the image analysis use case?
460.
by imjonse 3y ago
700 users on demo day. Too much writing code is bad for you.
461.
by imjonse 3y ago
SimCLR and others are older than 2021, BYOL is even mentioned in the blogpost. But your link indeed points to a more comprehensive overview.
462.
by imjonse 3y ago
When reporting fines for large companies in media, these should also be expressed as percentage of daily/yearly profits or revenues, to highlight the fact that most of the time they won't have any effect.
463.
by imjonse 3y ago
From the article I understand it's only for H100? Both the title and the charts imply that.
464.
by imjonse 3y ago
no code or dataset either for phi-1.
465.
by imjonse 3y ago
Congrats on the release! Two questions. 1) In the results table, Llama2 base is being compared to Persimmon base and finetuned, and only the latter performs better. Would a comparison to Llama2-chat be possible/fair? 2) The Llama-2 num
466.
by imjonse 3y ago
With at least 2x80G A100 needed for QLoRA finetuning and the 4bit quantized model requiring 90G only for the weights at inference time, I doubt this will put a dent in Llamas popularity outside large labs and enterprises. It may encourage m
467.
by imjonse 3y ago
I found the title misleading. Isn't learning from a single example desirable, while memorizing undesirable in the context of training? The former is the goal we're aiming for in order to match how animals learn, while the latter a
468.
by imjonse 3y ago
From the FAQ: ' Why would pretraining a 1.1B model for so long make sense? Doesn't it contradict the Chinchilla Scaling Law? Above is the training loss curve taken from the Llama 2 paper. Here I quote from that paper: "We obs
469.
by imjonse 3y ago
looked for this in the article, since it seemed such an obvious thing to test for but they haven't (yet). Vegan and vegetarian can mean both the menu and the people following that diet, "plant based" is clearly only about the
470.
by imjonse 3y ago
I wonder how useful these models are going to be for "Tracking Everybody Everywhere All at Once".
471.
by imjonse 3y ago
the author has been a respected Linux developer for decades, so he is probably familiar with that option.
472.
by imjonse 3y ago
If you look deep enough they are not exclusive. From the deeplearning book: "Many authors use the term “cross-entropy” to identify speciﬁcally the negative log-likelihood of a Bernoulli or softmax distribution, but that is a misnomer.
473.
by imjonse 3y ago
One may be tempted to use these findings as a justification for (indirectly) killing sentient mammals. Add studies showing that even plants and fungi have senses and try to avoid harm, and you might as well consider that there should be no
474.
by imjonse 3y ago
"If death is instantaneous, such as when you slap the mosquito on your skin, there is little room for suffering." This assumes that mosquitos perceive time as we do. While all living things operate with the same molecular building
475.
by imjonse 3y ago
The body's protein requirements are usually overestimated based on hearsay and this results in too much importance being given to animal based food. Unless you're a growing child, pregnant, bodybuilder or have a medical condition
476.
by imjonse 3y ago
Maybe they knew they were being spied on by Stalin. https://www.theguardian.com/books/2020/oct/11/revealed-sovie... Animal Farm is a great and easy read, but for something much less metaphorical, Homage
477.
by imjonse 3y ago
I have no experience with other AI assistants, but Cody for IntelliJ is free and it's not bad. It also uses cloud AI though, Anthropic Claude I think.
478.
by imjonse 3y ago
> And this is a superficial difference carried from old days when we need to do deployment and deployment-specific optimizations. Is it? From what I understand, to use an analogy, ONNX is the bytecode specification and JVM whereas Pytorc
479.
by imjonse 3y ago
your comment may have triggered a sentiment analysis model causing your accounts here and elsewhere to be flagged just in case they need to be 702ed in the future :)
480.
by imjonse 3y ago
A quote from that article "Transport is a small contributor to emissions. For most food products, it accounts for less than 10%, and it’s much smaller for the largest GHG emitters. In beef from beef herds, it’s 0.5%."

More ›