lxe

44 ms·

631.
by lxe 4y ago
I'm running 4-bit quantized llamas on torch/cuda with https://github.com/qwopqwop200/GPTQ-for-LLaMa , and I'm seeing significant tokens/second perf degradation compared to 8-bit bitsandbytes mode. I&
632.
by lxe 4y ago
It's just semantics. "managing remote resources" or "calling remote procedures" is all just network calls, usually http.
633.
by lxe 4y ago
There needs to be a carve-out for open source software. The difference between open source software and other industries is that OSS is not "the bridge" -- it's the "blueprint for the bridge". Until it's in the
634.
by lxe 4y ago
Looking at the control manual this looks very ergonomic. You're essentially delegating non-input controls to a joystick with feedback, while the typing is done on the keyboard. This is actually useful.
635.
by lxe 4y ago
I don't see anything wrong with this. It's like saying "Apple is built by Steve Jobs" even though obviously a lot more people are involved, and even if Jobs is no longer even around.
636.
Student falsely accused of using ChatGPT on high school paper(old.reddit.com)
3 points by lxe 4y ago|1 comments
637.
by lxe 4y ago
Thought it was interesting that we're seeing false positives and false negatives in these detection systems. > One of my papers from last semester before ChatGPT is shown to be 80% AI.
638.
by lxe 4y ago
> Therefore consumer applications will need to perform deduplication or handle incoming messages in an idempotent manner. > The way we achieve exactly-once delivery in practice is by faking it. Either the messages themselves should be
639.
by lxe 4y ago
What are you (they?) using for text to speech? Elevenlabs? Azure TTS?
640.
by lxe 4y ago
Some of these prompts sound like they were written by 14 year olds. But hey if it works, it works!
641.
by lxe 4y ago
There are still "hard" problems to solve when it comes to spellchecking. For example predictive text and correct typo detection. Notice what suggestions the Firefox and chrome spellcheckers give you versus what happens when you us
642.
by lxe 4y ago
Fun fact of the day: ffmpeg is written by Fabrice Bellard, who among other impressive things, wrote the JS PC emulator capable of running windows 2000 in a browser over 10 years ago that got me fascinated with emulators in general. https:&
643.
by lxe 4y ago
This is a collection of publicly available TV streams from all over the world. What could possibly be there that violates DMCA?
644.
by lxe 4y ago
Feels like whoever is running or judging entries for the TSA Webmaster Competition is not very literate in the field they are judging, if they consider Github a template engine.
645.
by lxe 4y ago
Yes! Love the vertical scaling approach. Pinning similar builds to same host doesn't get enough love.
646.
by lxe 4y ago
Awesome. How does this work in a distributed sense? For example, for multiple parallel builds where each build is building a different git sha?
647.
by lxe 4y ago
> Building Docker images in CI today is slow. CI runners are ephemeral, so they must save and load the cache for every build. >...persistent disks significantly lowers build time Does this mean your solution places specific caches, li
648.
by lxe 4y ago
Loving the opt-30b right now: Human: Can you write a song about the sun? Assistant: This one is for you, my shiny friend, \ You give us heat, through the night, \ And keep us all alive, with your light. \ You give us color, \ But most
649.
by lxe 4y ago
What's the best way to get started learning this? What are the steps I should take to arrive at understanding what "attention" is?
650.
by lxe 4y ago
Not much. Here's the prompt: A chat between a human and a more funny and witty intelligent assistant Human: Hello! What can you do? Assistant: Hey there! I can pretty much do whatever you ask! Human: What is the name of the t
651.
by lxe 4y ago
Got the ops-30b chatbot running on 3090 24GB. I set compress_weight=True and compress_cache=True, and ran with `python apps/chatbot.py --model facebook/opt-30b --percent 100 0 100 0 100 0`. I also modified the prompt a bit to make
652.
by lxe 4y ago
With compression, was able to get 30b to run on 3090 with '100 0'! Let me see if I can tweak the prompt a bit and make it come alive...
653.
by lxe 4y ago
Well just got some more sticks. While I wait for RAM to arrive, will try with compress_weight and compress_cache. If you're in any discord or any other space where people are tinkering with this, would love to join!
654.
by lxe 4y ago
How much system RAM are you running with? And I'm guessing it wouldn't hurt to have a fast SSD for disk offloading?
655.
by lxe 4y ago
Got the ops-6.7b chatbot running on a windows machine with a 3090 in mere minutes. The only difference was to install the cuda pytorch `pip install torch==1.13.1+cu117 --extra-index-url https://download.pytorch.org/whl&#x2F
656.
by lxe 4y ago
Hmm I thought this was going to be removing something profoundly out of touch or racist, but instead it's weird minor revisionism, like changing Kipling to Steinbeck for some reason, and updating various benign phrases with their moder
657.
by lxe 4y ago
It costs $30k to get exterior sewer main line replaced in many places in CA. I can get a simple iPhone app made on Fiverr for $300 This comparison makes no sense.
658.
by lxe 4y ago
The overarching issue is conflating error messages that are useful for engineers doing the debugging, like "token missing" and the cutesy ones sent to the user "oopsie whoopsie, computer broke :(". You can then use whate
659.
by lxe 4y ago
What would you say if you were in Russia? Would you publicly pour out your complete thoughts, whatever they may be, in a popular blog post with your name signed underneath? With an unfunded OSS project, a family to support and criminal hist
660.
by lxe 4y ago
A good take away about modern SBC landscape for hobbyists: you can get much better specced "mini" x86 PC at the cost of slightly more power consumption, but with benefits of much better OS and hardware compatibility at a similar p

More ›