segmondy

22 ms·

661.
by segmondy 2y ago
I did zero shot 27 solutions successfully with local model code-qwen2.5-32b. I think adding sonnet or latest gemini will probably get me to 40.
662.
by segmondy 2y ago
This is very wrong. Smaller models would generate response instantly on 4090. I run 3090's and easily get 30-50 tokens/seconds with small models. Folks with 4090 will see easily 80 tokens/sec for 7-8b model in Q8 and prob
663.
by segmondy 2y ago
Exactly, I'll personally buy it for $2Million for anyone that can get it and assign me the full code/weight and rights.
664.
by segmondy 2y ago
ASML has no competition, you will have an easier time replacing TSMC before doing ASML. Frankly, I think they should do a hybrid approach, have TSMC do some stuff and keep refining their own Fab till they figure it out.
665.
by segmondy 2y ago
they and TSMC don't compete with ASML. If you have the time, read "Chip wars" that will definitely convince you that USA needs a local chip maker and a Fab if the wish to retain their superpower position. Prior to reading t
666.
by segmondy 2y ago
Funny to see this, I bought 2 used DVD complete series from ebay last night instead of subscribing for it on prime even tho prime was a bit cheaper.
667.
by segmondy 2y ago
The board wants to split the company, the CEO didn't. That's what the fight was about. Should Intel become a design company? And have TSMC do everything? IMHO, nope. The rot in Intel happened a long time ago, the smart peop
668.
by segmondy 2y ago
Strange, any phone today would be better than most hardware we had 30 yrs ago. In the 90's I had monitors with resolutions of 320x240, limited info on the screen. Things were slow. If you gave me the option of using a phone from to
669.
by segmondy 2y ago
They don't need to do 128gb, 48gb+ would eat their lunch, Intel and AMD are sleeping.
670.
by segmondy 2y ago
Intern | Remote I'm looking for an intern to help me with my LLM experiments. You are into LLMs and generative AI if not obsessed and are already experimenting with them on your own. You use them almost daily and have used them via A
671.
by segmondy 2y ago
It's ridiculous how expensive the wrong hire can be https://www.wired.com/story/google-brain-ai-researcher-fired...
672.
by segmondy 2y ago
So how did we grow these decades ago when when we didn't have this fancy irrigation systems?
673.
by segmondy 2y ago
Great write up Bill!
674.
by segmondy 2y ago
So they want an open protocol, and instead of say collaborating with other people that provide models like Google, Microsoft, Mistral, Cohere and the opensource community, they collaborate with an editor team. Quite the protocol. Why shoul
675.
by segmondy 2y ago
RPC for LLMs with the first client being Claude Desktop. ;-)
676.
by segmondy 2y ago
How to sit down still and write lots of programs. How to sit down still and read lots of programs.
677.
by segmondy 2y ago
OpenAI is borrowing lots of ideas out there as they search for ways to make money. If you have a github project then they can borrow your idea. However if you have an idea that's closed source or behind an API, your idea could be ad
678.
by segmondy 2y ago
A lot of old stuff are still very relevant, they are much resource efficient, smaller amount of RAM, CPU, performance that's much faster. Easily 100x cheaper. I mean, if you have a few things you are trying to sort out, then using L
679.
by segmondy 2y ago
... and this is why those of that are into local models are into it.
680.
by segmondy 2y ago
Nah, I think OP has reached his idea of pinnacle. When you set out to do a thing, and achieve that thing, it's all downhill unless you set out for more challenging goals or switch up. People often get depressed, after getting their
681.
by segmondy 2y ago
Need more memory, 256GB will be nice. MistralLarge is 123B. Can't even give a quantized Llama405B a drive. LLM users rejoice. LLM power users, weep.
682.
by segmondy 2y ago
I think the big thing the skeptics miss is that they believe it's only good for wealth loss. But studies is showing that it has massive health benefits that you don't get from dieting alone. Yes, losing weight by workout/d
683.
by segmondy 2y ago
Very nice. Any video of the cars? I
684.
by segmondy 2y ago
LLM experiments.
685.
by segmondy 2y ago
When is the price set? When I pick it up the time or when I check out? Would it be legal if I pick up an item that's marked $1 and at checkout it's now marked $5? What if I hold said item and give it to someone else, what woul
686.
by segmondy 2y ago
You can always update the code to use that. Meta releasing stuff on github is not trying to release the "bet" but to give a proof of concept. The licenses of those TTS system matters, it's not enough to be open. If this
687.
by segmondy 2y ago
why does it have to compete? can you get utilization out of it? my desktop is 12yrs old, i5 quad core. my thinkpad is 9 years old, my macbook is 10yrs old. so yes, a 5 yrs old laptop is more than sufficient for you. I find that for lap
688.
by segmondy 2y ago
I was around before that wiki page was written and know exactly what open source means. I'm not going to play the double speak game on words. open source means you can read the code. period. it doesn't matter if it's free
689.
by segmondy 2y ago
semantics, "source available" is for closed software that's willing to make source available to their customers. For instance, Microsoft windows is not open source, but if you're an important enough customer, they will
690.
by segmondy 2y ago
It is open source, the source is right there on the repo for anyone to read. It can be open source without being free software, it can be open source without having any license. If the owner has opened their code to the world to read, it&#

More ›