zaptrem

19 ms·

211.
by zaptrem 2y ago
Need an updated version for those of us who were born after nearly everything mentioned here was no longer manufactured/supported/commonly used!
212.
by zaptrem 2y ago
Batch as many requests together as possible and your utilization will increase.
213.
by zaptrem 2y ago
Not op but my response: Because I am lazy and would like to save the 1-5 minutes it would take me to actually write it. When there are dozens of these small things a day the saved time really adds.
214.
by zaptrem 2y ago
I assume it knows the big stuff like the PyTorch API/major JS and React libs then just paste the docs or even impl code for any libs it needs to know beyond that.
215.
by zaptrem 2y ago
The alternatives are sadder: 1. We stick with the current wasteful and expensive models and have less money for everything else the government does. 2. We don't buy cruise missiles at all and Putin and friends gleefully march across ea
216.
by zaptrem 2y ago
We are but prediction machines https://www.psy.ox.ac.uk/news/the-brain-is-a-prediction-mach...
217.
by zaptrem 2y ago
It is irreversible similar to how a student reading a textbook from LibGen can remember and profit from that information forever. Kinda crazy how many in this community went from champions of freedom of knowledge to champions of megacorps o
218.
by zaptrem 2y ago
The Android running on your phone is not FOSS. If someone wants to write free software that runs on your phone they totally can (and do, check out Graphene).
219.
by zaptrem 2y ago
Would you pay for continued software support? IIRC Windows upgrades were $100-200 back in the day. More helpful answer: I think Graphene OS still has security updates for your model (assuming Google doesn't?)
220.
by zaptrem 2y ago
Deep faking voices has been real and super easy for years via 11labs etc. This doesn’t change that at all.
221.
by zaptrem 2y ago
For those of us using the 2D NATTEN kernel from their library along with torch.compile, is this faster? Especially given all their tricks (e.g., the non-deterministic KV-parallelism)
222.
by zaptrem 2y ago
I don't understand why all of their hubub is about tokens per second (which relies on a lot more than just hardware) instead of FLOPs/memory bandwidth/etc.
223.
by zaptrem 2y ago
I took the next level of that course and they taught up to SMT and a few other things. We did all of the homework assignments in a hardware description language called Verilog which allowed us to write abstractions for all the things.
224.
by zaptrem 2y ago
Oh, forgot about that one. Wikipedia says ReLU was used in NNs in 1969 but not widely until 2011. Idk if anyone has ever trained a transformer with sigmoid activations, but I don’t immediately see why it wouldn’t work?
225.
by zaptrem 2y ago
All of the main algorithms do (multi-layer perceptrons and stochastic gradient descent are from the 50s and 60s!). Basically the only thing that changed is we decided to multiply some of the outputs of the multi-layer perceptrons by each ot
226.
by zaptrem 2y ago
> Is FlashAttention simply a drop-in replacement for the attention operation in an LLM? Can it be used anywhere that an "attention" operation is used? Or does a LLM need to be trained specially to use FA? Yes > How does FA r
227.
by zaptrem 2y ago
Why did they give it such mediocre hardware and old software?
228.
by zaptrem 2y ago
Fabric is not a rare name https://lightning.ai/docs/fabric/stable/ https://github.com/FabricMC https://github.com/hyperledger/fabric
229.
by zaptrem 2y ago
I'm sure none of these people use freely available internet data as part of their jobs (i.e., for a profit), right? More specific to this situation, is there really much profit in geology chatbots?
230.
by zaptrem 2y ago
Likely this and radiologist practices and storage methods and patient demographics and a bunch of other things. However, if they're training these models with the intention of using them in other hospitals I'd still define that as
231.
by zaptrem 2y ago
Ah, that's fair. Didn't realize there was a significant performance impact.
232.
by zaptrem 2y ago
``` The researchers also found that they could retrain the models in a way that improves their fairness. However, their approach to "debiasing" worked best when the models were tested on the same types of patients on whom they wer
233.
by zaptrem 2y ago
// It seems that this just happens a lot in the wild. // We're seeing panics across various platforms where consecutive calls // to `Instant::now`, such as via the `elapsed` function, are panicking /&
234.
by zaptrem 2y ago
I interpreted it as “there’s such a big difference between artillery and a handgun you clearly have never been around an actual gun being fired before” Still snarky, but there is a point.
235.
by zaptrem 2y ago
This is already happening. At a previous job I saw an applicant who was struggling to figure out how to use any of the company’s standard tooling for version management and development because they spent their whole career only working with
236.
by zaptrem 2y ago
As a happy citizen of la-la land, what am I missing with desktop Linux (I use server Linux daily for ML and dev stuff) that’s worth all of the significant drawbacks that come with leaving la-la land?
237.
by zaptrem 2y ago
Why do y’all think the attacker was replying all of his requests? Could they be probing for unintentionally exposed endpoints themselves?
238.
by zaptrem 2y ago
There are other modern countries that have done worse to (and killed many more of) their own civilians.
239.
by zaptrem 2y ago
How exactly do you define atrocities? Nazi Germany and most ancient empires come to mind here.
240.
by zaptrem 2y ago
I think when CryTek died they gobbled up most/a lot of their engine engineering team and some crazy licensing. The Amazon Lumberyard "transition" seemed more like corporate partnership marketing to me considering how they&#x2

More ›