sanxiyn

24 ms·

361.
by sanxiyn 3y ago
Time-travel seems miraculous to me, but uptake seems very slow to me. I hope https://www.replay.io/ succeeds.
362.
by sanxiyn 3y ago
Being against a kernel debugger is one of the worst decision Linux ever made. Thankfully those days are behind us. https://lwn.net/Articles/270089/ Linus Torvalds, in paraphrase: > I happen to believe that ston
363.
by sanxiyn 3y ago
Safari has the fastest JavaScript engine. In many respects, Safari's implementation is top notch. Apple makes money off phones and people use Safari a lot on phones. I don't understand why people think Apple don't invest on S
364.
by sanxiyn 3y ago
Which Google stuff requires Chrome? I run even Google sites on Firefox and they seem to run fine.
365.
by sanxiyn 3y ago
> We have no control over it and there's nothing citizens can do to reduce our footprint. That's just not true. You can buy cars with better fuel efficiency. (Just say no to SUV.) You can buy electric cars. You can avoid air tr
366.
by sanxiyn 3y ago
We already do. Both are useful.
367.
by sanxiyn 3y ago
US and Canada drive larger cars with worse fuel efficiency. This is one of primary causes.
368.
by sanxiyn 3y ago
Individual consumption is in fact one of primary sources of emissions. That's not BP or Exxon propaganda, see what EPA says about it: https://www.epa.gov/ghgemissions/sources-greenhouse-gas-emis... Transportation
369.
by sanxiyn 3y ago
People did that too, and it is still decreasing after trade adjustment.
370.
by sanxiyn 3y ago
No. What US did is decarbonization and energy efficiency improvement. Trade-adjusted emission, taking into account of manufacturing outsourcing, is also decreasing, since 2005.
371.
by sanxiyn 3y ago
No it isn't. Consumption is not emission. We did improve energy efficiency a lot.
372.
by sanxiyn 3y ago
Consumption is not emission. We are in fact making much more efficient use of energy.
373.
by sanxiyn 3y ago
That doesn't change much. Trade-adjusted data is similar. US per capita emissions, both unadjusted and trade-adjusted, started to fall since 2005. https://ourworldindata.org/grapher/consumption-co2-per-capit...
374.
by sanxiyn 3y ago
All Debian packages are built natively. There is no cross build going on here.
375.
by sanxiyn 3y ago
If you want an unrestricted model, you should train one yourself. You don't want safety scissors, alas, we can't have all things we want, can we. Facebook is under no obligation to provide you one, after all it's Facebook&#x2
376.
by sanxiyn 3y ago
Only in EL1, which means you can only use them in kernel.
377.
by sanxiyn 3y ago
Yes. You can consider ggml akin to PyTorch, and llama.cpp like Transformers (by Hugging Face).
378.
by sanxiyn 3y ago
No. The C code can read in a model weight, take input, and return output, but it runs on CPU, not GPU. It also can't run any other models, unlike PyTorch. The model is hardcoded to Llama 2.
379.
by sanxiyn 3y ago
Debian packages declare their supported architectures. "all" means it is architecture independent and will be available to a new architecture without rebuild. "any" means it is portable and will be available to a new arc
380.
by sanxiyn 3y ago
Debian does distribute SSE-using binaries (with fallback) on i386 which detects presence of SSE at runtime using CPUID.
381.
by sanxiyn 3y ago
> Check for feature presence That is the hard part. On Intel you can use CPUID, but it is ARM policy to not expose such instructions. You can read /proc/cpuinfo, but that is Linux-specific. Edit: there is a reason for ARM polic
382.
by sanxiyn 3y ago
Currently the C code does not invoke Python and there is no way to pass a prompt. It does not need any special permissions.
383.
by sanxiyn 3y ago
Note that NEON is optional for armhf, mainly to support Tegra 2. This was a big deal at the time. See https://wiki.debian.org/ArmHardFloatPort .
384.
by sanxiyn 3y ago
Apple did LLVM and WebKit, which seem far more important than React and PyTorch.
385.
by sanxiyn 3y ago
"Forced" or not, they still raised rates. They could have done otherwise, which would have been disastrous.
386.
by sanxiyn 3y ago
Well, inflation is an actual danger (see Turkey) and Federal Reserve deserves credit for handling inflation well. I agree semiconductor is an industry well known to be cyclic and prophecies of doom are ridiculous.
387.
by sanxiyn 3y ago
Generally you need background knowledge to read papers. A quick way to gain necessary background knowledge is to read literature reviews or surveys. If you can't understand even reviews, surveys, or introductions, it is time to read te
388.
by sanxiyn 3y ago
Principles of Design (1998) by Tim Berners-Lee. Especially its Principle of Least Power. https://www.w3.org/DesignIssues/Principles.html
389.
by sanxiyn 3y ago
I already said he made many insane decisions. This is not one of them.
390.
by sanxiyn 3y ago
SBF made many insane decisions, but preparing for the end of the world seems entirely sane.

More ›