Kubuxu

8 ms·

61.
by Kubuxu 3y ago
Welcome in the SDAM (Severely Deficient Autobiographical Memory) club. It is highly correlated with Aphantasia.
62.
by Kubuxu 3y ago
1. Development started in 2000, and the first airframe (albeit with non-finalized parts) rolled off in 2006. 2. The pilot can make independent decisions even in the presence of overwhelming communication jamming.
63.
by Kubuxu 3y ago
The F-35 program was developed with the principle of concurrency in mind. The first aircraft rolled off the production line before the majority of the parts were finalized. Think of it as the first build from your CI/CD pipeline. The b
64.
by Kubuxu 3y ago
Some components of the car can be shared, others have to be shared, but break ducts IIRC are in the third category where you have to design them yourself.
65.
by Kubuxu 3y ago
We use some radio telescopes for radar astronomy. https://en.wikipedia.org/wiki/Radar_astronomy
66.
by Kubuxu 3y ago
I think of the SD3 as a further evolution of SD1.5/2/XL and StableCascade as a branching path. It is unclear which will be better in the long term, so why not cover both directions if they have the resources to do so?
67.
by Kubuxu 3y ago
I use zsh-histdb which stores the history in SQLite with context. Then I have few different queries/ways to interact with it. For example I have a line completion (which also shows as grayed out text while I type) based on most frequen
68.
by Kubuxu 3y ago
AFAIK it was released to "researchers" behind submitting a survey, someone filled out the survey and put it on torrents. There was no way it wouldn't leak.
69.
by Kubuxu 3y ago
You could consider something like BLIP2. There are multiple ways you could use it: embed images and match them against embeddings of text descriptors, train custom embedding of descriptors, or a classifier on top of the embedding (linear la
70.
by Kubuxu 3y ago
AFAIK the EU AI Act primarily prevents profiling, social scoring and other high risk activities from using AI. It generally didn’t care about generative AI.
71.
by Kubuxu 3y ago
Make sure to change omnibar settings to prefer history over search.
72.
by Kubuxu 3y ago
Which is why I have disabled that functionality. I also don’t need M$ to know every app I’m running on my PC (although they probably have that data some other way).
73.
by Kubuxu 3y ago
The 3rd paragraph missed the mark but previous ones are in the right ballpark. You take the users question either embed it directly or augment it for embedding (you can for example use LLM to extract keywords form the question), query the v
74.
by Kubuxu 3y ago
In essence batching allows for more efficient usage of memory bandwidth. Without batching for every generation you need to transfer the whole model from GPU memory to GPU core once for every image which sets an upper bound on speed. With ba
75.
by Kubuxu 3y ago
It is two out of eight at each layer, with 32 layers independent of each other. There are no eight "sub-models". However, this raises a question: could a slightly more complex router use output layer n-1 to choose experts for laye
76.
by Kubuxu 3y ago
Nothing prevents C&C from being fronted by cloudflare.
77.
by Kubuxu 3y ago
Very LLMs are quite a different from diffusion models. The model size vs training set size is skewed the other way.
78.
by Kubuxu 3y ago
Mozilla allows for direct update of blocker lists instead of requiring that they are included in the extension and updated with it.
79.
by Kubuxu 3y ago
With a fast NVME loading a model is only 2-3s.
80.
by Kubuxu 3y ago
ggml.cpp with blast backend could be one example of it See for instance: https://github.com/ggerganov/ggml/blob/57c468b8655f3630d1749... which are the parts not available in blast
81.
by Kubuxu 3y ago
No, it can't, as it primarily supports only domain blocking.
82.
by Kubuxu 3y ago
> you get true cross-platform support wasi_nn is just an abstraction layers for wrappers around llama.cpp/gguf.cpp/onyx/others. It is not any more cross-platform than these libraries/apps deployed themselves, or rust
83.
by Kubuxu 3y ago
They also weren't targeting orbit but just bellow it.
84.
by Kubuxu 3y ago
Your KV cache size is linear with context size which might put you tight on memory. There is also increased cost of recalculating KV cache of context window when the window has to move but this is close to being solved with streaming LLMs.
85.
by Kubuxu 3y ago
In most tanks fuel is used as higher density voidspace to decelerate projectiles and shrapnel.
86.
by Kubuxu 3y ago
I couldn’t agree with you more. Today when we are talk with LLM it’s like we (sorry for anthropomorphism) interact with a naked mind and we rely on pure immediate recall and it’s train of thought. I wonder whether allowing LLMs to perform i
87.
by Kubuxu 3y ago
Disable memory overclock or undervolt the memory and SoC. You will see the power drop down like crazy.
88.
by Kubuxu 4y ago
Enchanter for YouTube allows you to set the speed as fast as you want.
89.
by Kubuxu 4y ago
Yes, that is why residents are paid pennies. But once you are fully qualified surgeon, your are in the artificially limited pool of people who can operate on patients. Which leads to high salaries for surgeons.
90.
by Kubuxu 4y ago
You can fix it quite easily by pulling out the whole switch and setting it in the box the other way.

More ›