rfoo

42 ms·

601.
by rfoo 3y ago
Just the Internet. You have to click Skip when the phone prompts you to log in to a Google account though, as it does this by default.
602.
by rfoo 3y ago
The two reasons are invalid for Pixel phones: 1. Pixel phones display an "unlocked bootloader" warning (which can't be disabled) during boot. In case it is re-locked with an additional signing key installed (Pixel-s are liter
603.
by rfoo 3y ago
> above 128gb RAM at the same speeds or better No way! That would be too expensive for us! To be able to really "crush" Apple, vendors MUST charge the same amount of money for RAM upgrade, and 8GB -> 128GB would be $3000!
604.
by rfoo 3y ago
"How do you xor two numbers in Julia?" "a ⊻ b" "... am I supposed to type (or copy paste) that symbol every time I want to xor two numbers? There must be a better way right?" "Sorry, no." I immediatel
605.
by rfoo 3y ago
But the same also applies to C++. A boatload of my compile time is spent on expanding, parsing the same header files again and again in different compilation units. This can be solved by precompiled headers, yeah. But then again a boatload
606.
by rfoo 3y ago
Calling public (but undocumented) APIs without authentication is NOT exploiting bugs. If this is "exploiting bugs" then by the same logic I'm hacking the government website by clicking view source.
607.
by rfoo 3y ago
Hehe, you can solve the two "Street" challenges with GPT-4, they are bog standard entry-level competitive programming challenge disguised as a reverse engineering challenge.
608.
by rfoo 3y ago
Thanks!
609.
by rfoo 3y ago
Question: Have you considered one separate binary (but still sharing the entire codebase) for each "workload" instead of using a flag to switch? If so, can you highlight the advantage of the flag way?
610.
by rfoo 3y ago
> how else do you think malicious PDD app got on their phones? The malicious PDD app is really, actually, published to the alt stores by PDD Holdings itself. It loads the exploit config and post exploitation modules from PDD's own C
611.
by rfoo 4y ago
> The image encoder is implemented in PyTorch and requires a GPU for efficient inference. WebGPU just shipped today in Chrome, there are no people reporting that the demo doesn't work with their days-old browser, so it doesn't
612.
by rfoo 4y ago
Yes, FAQ said: > What platforms does the model use? > The image encoder is implemented in PyTorch and requires a GPU for efficient inference. > The prompt encoder and mask decoder can run directly with PyTroch or converted to ONNX
613.
by rfoo 4y ago
Thanks for sharing your experience. Yeah, that's my point, it's hardly a solved problem and you have to write software for this!
614.
by rfoo 4y ago
PCIe lanes: it's not that important since the data exchange between host and device during inference is pretty minimal. For model parallel / tensor parallel you want fast interconnect between GPUs, the ages old tech for this is PC
615.
by rfoo 4y ago
That's a viable solution, but since GPU instances are expensive, you really want to make most of it. Ideally the GPU should be busy within 30 seconds of instance launch. Okay, so, where is your training data? Is your training data in t
616.
by rfoo 4y ago
Yes? The original LLaMA-65B on fp16 weight is 130G. There are no NVIDIA GPU with 130G memory. As for why they don't use int8/int4: would you try to optimize your code before you confirmed it can actually run?
617.
by rfoo 4y ago
Congratulations on rediscovering how the original researchers run these models!
618.
by rfoo 4y ago
There's https://en.wikipedia.org/wiki/Tuanpai and https://en.wikipedia.org/wiki/Shanghai_clique . And of course a third one - Xi organized his own clan.
619.
by rfoo 4y ago
There's deeplearning4j (from Theano days!), go figure why it didn't take off.
620.
by rfoo 4y ago
Check https://pytorch.org/tutorials/advanced/cpp_frontend.html You can easily build a standalone binary (well, it would be GiB+ if you use CUDA... but that's the cost of statically linking cu*), had you coded
621.
by rfoo 4y ago
The stack is very volatile and unmaintainable because it doesn't need to be maintainable. Exactly why we have unmaintainable software in other domains. During the last 10 years there are ALWAYS totally new model architecture with new o
622.
by rfoo 4y ago
Sigh. It's not like the zero copy buzzword is going to help you during training, all your weights have to stay on GPU, you are going to sample your training data randomly and your data is on a networked storage anyway, so mmap HURTS. Y
623.
by rfoo 4y ago
The real Azure and AWS is accessible in China, too. In fact, that's the easiest reliable way to bootstrap from "I'm in China and have no censorship-circumvent tools" to accessing the real Internet. Just go to Azure or AW
624.
by rfoo 4y ago
> and that's enough to meet their main goal, protecting domestic tech companies from competing with established US tech companies. You haven't experienced the GFW if you think that's the main goal. Oh, and please try telli
625.
by rfoo 4y ago
Previous discussion: https://news.ycombinator.com/item?id=35269347
626.
by rfoo 4y ago
A few dozen people arrested won't make me think it's the worst. But how about a few student activitists disappeared, beaten to confess in a video clip, and two of them still missing even today? From the Wikipedia: > In January
627.
by rfoo 4y ago
TBF at least half of the firms did't give a fuck to the specific regulation at that time, and given the rumor that the bug is found when a Security Engineer (who works on product security instead of vulnerability research) decided to l
628.
by rfoo 4y ago
China ran the worst brutal crackdown against working class protesters (and students trying to help) in the recent years: https://en.wikipedia.org/wiki/Jasic_incident
629.
by rfoo 4y ago
There's SSH agent restriction now. [1] https://www.openssh.com/agent-restrict.html
630.
by rfoo 4y ago
Not GP, but: I do, however when I do this I make sure the certificate is signed with permit-agent-forwarding and demand people just forward their ssh agent on their laptops. This also discourages people from leaving their SSH private key on

More ›