namibj

20 ms·

331.
by namibj 1y ago
Worse: the 16€ option only curtailed in tracking& insurance, along with a secret height limit (35x25x3 cm limit), is now unavailable for shipping say a singular music CD, or an issue of a magazine. Sure, paying on the order of the goods
332.
by namibj 1y ago
(on the order of) 110 bits of brute-force isn't anywhere close to "a decade" especially if you aren't throwing severe parallelism at it. Edit: we're talking on the order of a million times the current age of the uni
333.
by namibj 1y ago
Or, instead of "estimate the distribution" look at what conceptual properties of the instance the big-O cares about. And then derive big-O in terms of such properties. See for example Tetris-LoadBalanced and Tetris-Reordered (on t
334.
by namibj 1y ago
Arguably this is just a fairly poor way of thinking for quite many practical applications. Notably, memory access latency(/inverse throughput) is roughly between O(log(n)) [ https://news.ycombinator.com/item?id=12385472
335.
by namibj 1y ago
Watching rented movies is something consumers apparently do; I understand it to have mostly been used for that. Besides that, these aren't area-heavy features; it's cheaper to share the core design and just have the feature availa
336.
by namibj 1y ago
Move to console or an arcade or something, but away from general purpose personal computers, if you want that level of assurance from the system. Please. They'll take our collective computing freedom if we don't keep these separa
337.
by namibj 1y ago
The problem is that the ring requires suitably locked memory which won't be subject to swapping, thus inherently forcing a different memory type if you want the extra-low-overhead/extra-scalable operation. It makes sense to ask th
338.
by namibj 1y ago
Such reads are in principle supported if you have sufficient hardware offloading of your stream. AFAIK io_uring got an update a while back specifically to make this practical for non-stream reads, where you basically provide a slab allocato
339.
by namibj 1y ago
It's from the loco which in the US almost exclusively used electric propulsion, just for capex vs. opex balance sheet gaming reasons mostly (except in and around NYC (tunnels) and some very recent electrification efforts (I think brigh
340.
by namibj 1y ago
Go back and edit a prompt of yours in the conversation instead of continuing with garbage in the context.
341.
by namibj 1y ago
Bold of you to assume Walmart and the like train their cashier's on speed. (I wish I was kidding; discounters that squeeze costs everywhere including cashier throughput seem to be the exception in retail.)
342.
by namibj 1y ago
Do TPUs allow having a variable array dimension at somewhat inner nesting level of the loop structure yet? Like, where you load expensive (bandwidth-heavy) data in from HBM, process a variable-length array with this, then stow away/acc
343.
by namibj 1y ago
That's about local link loss; you at best get buffer bloat from confusing the wired desktop and the wireless laptop that share a 800~1200 Mbit/s DOCSIS downlink. Or worse, different service tiers from a neighborhood getting bundle
344.
by namibj 1y ago
IIUC volta brought the ability to run a tail call state machine with let's presume identically-expensive states and state count less than threads-per-warp, at an average goodput of more than one thread actually active. Before it would
345.
by namibj 1y ago
L4S[0] also helps a lot with sensing congestion before young connections have yet to suffer their first lost packet... Basically it sqrt's your actual packet loss rate as far as feedback frequency/density is concerned, without act
346.
by namibj 1y ago
You're forgetting wireless access networks with varying signal strength (home wifi; phone in the basement).
347.
by namibj 1y ago
lora hyper parameter change? Defaults may not be tuned for knowledge insertion , but rather for style imprinting.
348.
by namibj 1y ago
Basically re-ranking is similar to how RAG functions for normal LLM tasks, in that you smash the fragment down into an embedding vector you use for a vector database to coarsely search, and then you use a powerful model with the task/q
349.
by namibj 1y ago
Cables much better sadly, so much so that they started to use cables to jump across the server main board in places.
350.
by namibj 1y ago
Those exist; they are used for risers ("vertical mount GPU brackets, for dual GPU" equivalent for servers, where they make the cards flat again).
351.
by namibj 1y ago
There are 4 major global coverage ones plus the JP one (QZSS) and iirc also some recent Indian. The major ones are GPS, Galileo, Glonass, Baidou.
352.
by namibj 1y ago
Yes it is, GPS is 1foot ~ 1ns; thus a few ns are trivial with a vaguely decent receiver.
353.
by namibj 1y ago
But we can regrow them. We just evolved an anti-teeth-regrowth substance/molecule that's in our blood and shuts down teeth growth once adult teeth are finished, because adult teeth roots are so deep that they require surgery to pu
354.
by namibj 1y ago
Doesn't need to be a heatpump dryer, just needs to be ventless. E.g. the 20+ year old common German dryer that runs the clothes air closed-loop, first through the clothes, then to a heat exchanger with drainage geometry (typically feed
355.
by namibj 1y ago
Just wanna remind that transformers are devoid of any "ordering/sequence" concept until you feed one in via positional encoding. It'd be easy to flag retracted tokens as such (e.g. pointing an input token one direction o
356.
by namibj 1y ago
I specifically mean prompts here, and I don't mean they'd have casual attention. Just run an encoder to get your KV cache pre-filling of the prompt, then do non-causal diffusion generation of the response referencing the cached pr
357.
by namibj 1y ago
You need a retraction/correction mechanism so the diffusion isn't locked in on a bad choice in order to really reduce iteration count, sadly.
358.
by namibj 1y ago
I recommend using optical networking if you are confident about the lightning strike Ethernet damage.
359.
by namibj 1y ago
Even my old GTX 970 can throttle down to like 10W while still being able to display and iirc also h.264 decode 1080p60, let alone putting it in a mode that at all matches S3/suspend-to-ram via PCIe sleep states. I'm pretty sure la
360.
by namibj 1y ago
Or buy equipment targeting alirack compatibility, i.e., with 240V PSUs that are minimal effort designed to also specifically work on the appropriate DC voltage so that Alibaba, the alirack standard's originator, could delete the invert

More ›