Hacker News new | past | comments | ask | show | jobs | submit | 1R053's comments login

one of the annoying things about Mathematica is that all functions are crammed into the same namespace and that there is no overloading with different parameterization options...


What do you mean by overloading? Functions can easily have different behavior with different argument counts (e.g., 2- vs. 3-argument Fold), and they can have any number of options (e.g., Graphics, Graphics3D, Solve, Import/Export, etc.). The only big redundancy I can think of is the various Plot functions.


creative destruction is not necessarily bad, but the default way of nature to move forward.

Obviously, you need to destroy the right things...


To everyone voting it was clear, that Elon would get this role. So he was part of the elected package. Probably, that was also a deciding factor for many voters.


An H100 supports 80 GB of memory. so at FP8 that would allow 3 of the 16+1 models per GPU (assuming around 26B per model), requiring 9 H100s, that usually would not fit one chassis I guess.

Once you have something with 192 GB it gets interesting. You could probably have 7 at FP8 per GPU. At FP16 it probably only would fit 3 per card, requiring 9 again.

I'd say for the current memory layout of cards they missed a little bit the sweet spot. With slightly smaller models or one expert less one should be able to run it on 8 H100s at FP8 or 2 B100s at FP8 or even on 4 B100s at FP16 if I calculated correctly.


You could always split one of the experts up across multiple GPUs. I tend to agree with your sentiment, I think researchers in this space tend to not optimize that well for inference deployment scenarios. To be fair, there is a lot of different ways to deploy something, and a lot of quantization techniques and parameters.


you can get a long way on something with 41% less performance than your favorite supercar...


the paper with details: https://arxiv.org/pdf/2411.02265

They use

- 16 experts, of which one is activated per token

- 1 shared expert that is always active

in summary that makes around 52B active parameters per token instead of the 405B of LLama3.1.


how fast do these machineries move? as I understand the weather scenario the wind direction was changing rapidly by more than 90 degrees at hurricane level. That means, the boat was unable even under power to redirect into the wind timely enough. The key question is, how much side wind can the design withstand without getting pressed into more than 45 degrees.

obviously lowering the keel affects this a lot, as there is much difference in leveraged 60 tons vs non-leveraged.


say moving 100tons 10m, like from one side to another, in 10 seconds - that, acceleration and deceleration, takes 200KWt - half the Tesla's electric motor. It is horizontal. As we may have heeling, the worst case - raising 100t vertically 10 m in 10 sec - 1000KWt, 2-3 Tesla's electric motors.


probably the next update wave is coming from the need of AI features for more local memory and compute. The software is just not there yet in usual tasks but it's just a question of time I guess. Of course there will be the pressure to do that in the cloud as usual, but local compute will always remain a market.

and probably it's good that at least one of the big players has a business model that supports driving that forward


I do empathise with your desire a lot

...but then such a service gets monetised by advertising, financed with VC money and guess what happens next...


it must suck to be that cynical


*unable/unwilling to self-deceive ;)


Can you recommend a good tutorial?


The first step might actually be a non-starter for most people. Notion is free, SharePoint is usually offered as a capability within a business or school's Office 365 subscription. I believe you can either subscribe yourself for $5.00 / month using this link: https://www.microsoft.com/en-us/microsoft-365/sharepoint/com..., or maybe you're lucky and already have an account with your school / work. If neither of these work for you, you can also sign up for the Microsoft Developer Program, which is actually pretty cool; here is a guide for setting up a SharePoint site through the developer program: https://www.sharepointdiary.com/2021/05/how-to-create-sharep...

I should take a moment to clarify that I'm not affiliated with Microsoft or any of the links I'm posting here in any way. I make a living building web sites and as mentioned above, many of them leverage SharePoint. I find it a pretty easy way to build intranet solutions for my customers.

Assuming you've got your SharePoint site created within SharePoint Online, creating a "List" is pretty simple. Here's an article from Microsoft explaining how to do so: https://support.microsoft.com/en-us/office/create-a-list-0d3.... Think of a list as a database table. From there you can create various types of columns including "look-up" columns, which are basically joining to another table on the same site with a foreign key.

I think from there you could dive into the Power Platform, but there's so much out there, I think this would be a good start.


I will try to gather some resources later tonight. If needed I’ll type up something more comprehensive.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: