jart

44 ms·

871.
by jart 3y ago
Isn't most of science like that too?
872.
by jart 3y ago
Scratchpad lives on in FriCAS https://github.com/fricas/fricas/commits/master/ which has commits 18 hours ago. Tim Daly worked diligently on Axiom for years. However some of his changes, like requiring e
873.
by jart 3y ago
What does red tape have to do with autonomy? Lack of autonomy means you work at the kind of place where you'll get a dressing down for fixing a GitHub issue that wasn't specifically assigned to you by your product manager. That&#x
874.
by jart 3y ago
Thanks for linking that. It was a nice half an hour well spent.
875.
by jart 3y ago
That's because Google never went mask off. When Google started, only nerds cared about the Internet. As the Internet became more successful, non-nerds started demanding that the company do for them all the evil things non-nerd society
876.
by jart 3y ago
It's here! Download the nightly build, which is a fat binary for x86+arm: wget https://cosmo.zip/pub/cosmos/bin/redbean chmod +x redbean ./redbean Enjoy! I'll get a release out o
877.
by jart 3y ago
Only the NetBSD community could have come up with a project this cool!
878.
by jart 3y ago
> Rust is really interesting for stuff besides running the actual llm though. That's the reason why llama.cpp is so attractive though. It just runs the LLM.
879.
by jart 3y ago
The greatest of all quiche eaters has just passed away. May he rest in peace. https://www.pbm.com/~lindahl/real.programmers.html But seriously, PASCAL was the first programming language I loved that's actually goo
880.
by jart 3y ago
Do you know how much faster llama.cpp would go on something like an Intel Core i9 (has AVX2 but not AVX512) when it's compiled using `cmake .. -DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=Intel10_64lp -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER
881.
by jart 3y ago
We're working on removing the hard-coded /bin/foo paths from the ape bootloader. Just give us time. Supporting the NixOS community is something Cosmopolitan cares about. Until then, try using the ape binfmt_misc interpreter.
882.
by jart 3y ago
I'm not exactly building bridges or x-ray machines, or putting food in anyone's mouth. Right now I'm building programs that generate text, and it isn't even real text like in a book that could potentially fall off a shel
883.
by jart 3y ago
TCO is cool. I remember how enlightening it felt when I studied Scheme and came to understand how it's implemented there.
884.
by jart 3y ago
Are you saying that knowing LISP makes you like an instant Paul Graham? If only.
885.
by jart 3y ago
> I'm sure this, and other LLM/IDE integration has it's uses, but I'm failing to see how it's really any kind of major productivity boost for normal coding. The Duomo in Florence took multiple generations to buil
886.
by jart 3y ago
If you just run this, without Emacs: ./wizardcoder-python-34b-v1.0.Q5_K_M.llamafile Then it'll launch the llama.cpp server and open a tab in your browser.
887.
by jart 3y ago
On Asahi Linux you might need to install our binfmt_misc interpreter: sudo wget -O /usr/bin/ape https://cosmo.zip/pub/cosmos/bin/ape-$(uname -m).elf sudo chmod +x /usr/bin&#x2
888.
by jart 3y ago
On Silicon I can guarantee that the Cosmo Libc emacs prebuilt binary will have zero trouble launching a llamafile process. https://cosmo.zip/pub/cosmos/bin/emacs You can also edit the `call-process` call so i
889.
by jart 3y ago
Maybe one day, when I have the resources to get into training, I'll do something like that in order to create ChatJustine :-) Until then I like to keep the technique behind my code craft private, which is one of many reasons why I love
890.
by jart 3y ago
You can download the llamafile executables from Mozilla's release page here: https://github.com/Mozilla-Ocho/llamafile/releases and then use the `-m` flag which lets you load any GGUF weights you want from Hu
891.
by jart 3y ago
llamafile is designed with the hope of being a permanently working artifact where upgrades are optional. You can upgrade to new llamafile releases in two ways. The first, is you can redownload the full weights I re-upload to Hugging Face wi
892.
by jart 3y ago
You're the third person in the last 40 minutes to post a comment in this thread sharing a link to promote that project. https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que... It must be a good proj
893.
by jart 3y ago
I was talking about showing empathy. I'm not sure where you got those ideas.
894.
by jart 3y ago
That's because learning how to use Emacs is basically the equivalent of Navy Seals training except for programmers. For Emacs believers, that's a feature, not a bug. The good news is that llamafile is designed to be easy and inclu
895.
by jart 3y ago
Both the Emacs Copilot and the Mozilla Ocho READMEs link to the canonical source where I upload LLMs which is here: https://huggingface.co/jartine/wizardcoder-13b-python/tree/m...
896.
by jart 3y ago
Try downloading https://cosmo.zip/pub/cosmos/bin/assimilate-x86_64.macho chmod +x'ing it and running `./assimilate-x86_64.macho foo.llamafile` to turn it into a native binary. It's strange that
897.
by jart 3y ago
I don't see why not. You'd probably just change this code: (with-local-quit (call-process "wizardcoder-python-34b-v1.0.Q5_K_M.llamafile" nil (list (current-buffer) nil) t
898.
by jart 3y ago
Fair enough. Myself on the other hand, I want the LLM to think when I tell it to think (by pressing the completion keystroke) and I want to be able to supervise it while it's thinking, and edit out any generated prompt content I dislik
899.
by jart 3y ago
If Microsoft is unhappy with 70 lines of LISP that I posted on their website, then I'm more than happy to change it. Ask them to reach out to jtunney@gmail.com
900.
by jart 3y ago
Thanks!

More ›