M@ · @Mrw
15 followers · 59 posts · Server hachyderm.io

Why am I supposed to be afraid of taking my job?

My current job is figuring out the exact set of instructions that will cause a computer program, be it the controller or or the compiler, to do what I want it to do.

If I try to imagine a dystopia where has taken my job, and I am relegated to figuring out the exact set of instructions that will cause a computer program, in this case an LLM, to do what I want it to do, well…gosh.

What a nightmare.

#ai #k8s #zsh #golang #codellama

Last updated 1 year ago

mau 🏳️‍🌈#PhaseOutFossilFuels · @mzedp
854 followers · 2655 posts · Server mas.to

Well, it managed the first problem easily.

#projecteuler #llm #llama #codellama

Last updated 1 year ago

mau 🏳️‍🌈#PhaseOutFossilFuels · @mzedp
854 followers · 2635 posts · Server mas.to

Ok, so, to be clear, the things I'm asking CodeLlama to do are not complex - but they ARE the kind of things I often need to do with programming, and find myself searching over and over for.

So... Having a program running on my PC that I can just ASK to give me these answers is extremely valuable to me.

LLAMA AI

#llm #codellama

Last updated 1 year ago

mau 🏳️‍🌈#PhaseOutFossilFuels · @mzedp
854 followers · 2632 posts · Server mas.to

For instance, you can't "ask" it "write this program for me" - BUT you CAN provide a comment describing what the program is meant to do, and then let the AI autocomplete the rest of the program.

That seems to work pretty well!

#ai #llm #llama #codellama

Last updated 1 year ago

mau 🏳️‍🌈#PhaseOutFossilFuels · @mzedp
852 followers · 2626 posts · Server mas.to

I considered subscribing to ChatGPT.

Having access to a much better and faster programmer than myself, on call 24/7, to help me with my projects, for 20$ a month, seemed like quite a bargain.

But... Meta just released an Open Source AI model which apparently outperforms ChatGPT in programming, and I can run it locally.

So... Let's see what this CodeLlama can do.

#llama #codellama #llm #ai

Last updated 1 year ago

Jan Eggers · @janeggers
556 followers · 632 posts · Server hessen.social

Frohe Kunde für -Nutzende: läuft!

#ollama #codellama #ki #llm

Last updated 1 year ago

· @MrClon
250 followers · 11866 posts · Server lor.sh

Is there some thePyrateBay for ? Want to test but Meta require some jumps over hoop to download it

#neuralnetworks #codellama #llm

Last updated 1 year ago

FOSSlife · @fosslife
2054 followers · 223 posts · Server fosstodon.org
Benjamin Han · @BenjaminHan
460 followers · 1263 posts · Server sigmoid.social

2/ Since users of with particular APIs are usually relatively inexperienced in the said APIs, these inaccuracies may have grave consequences to the robustness and reliability of the resulting software.

(How would fare?)

#codegeneration #codellama #generativeAI #papers #nlp #nlproc #softwaredevelopment

Last updated 1 year ago

Mark Pesce · @mpesce
1526 followers · 2542 posts · Server arvr.social

I am not a programmer but I suspect this code sample is wrong (via Instruct 34B model, running very comfortably on my 32GB MBP with 4-bit quantisation)

Interesting to note it will not generate COBOL (sorry, bankers) and refuses to believe FORTH has arithmetic operators.

And really, all FORTH has is stack and arithmetic operators. 🧐

#lua #codellama

Last updated 1 year ago

jfk · @jfkimmes
103 followers · 92 posts · Server social.tinycyber.space

Finetuned code-llama beats GPT-4. They have no moat, indeed.

phind.com/blog/code-llama-beat

#llama #codellama #gpt4 #ai #chatgpt

Last updated 1 year ago

Rob · @Feynman
116 followers · 83 posts · Server mastodon.sdf.org
Mark Pesce · @mpesce
1526 followers · 2533 posts · Server arvr.social

It looks as though the community is making a meal of Meta's only-yesterday-released .

One group is now claiming (and they've released everything so they expect others to be able to support this claim) that they're scoring higher than GPT-4 with their fine-tuned CodeLlama model.

Built in less than 24 hours after the release of CodeLlama.

And huge if true, because this model, with 34 billion parameters, can run on a well-equipped PC.

phind.com/blog/code-llama-beat

#foss #ai #codellama

Last updated 1 year ago

IT News · @itnewsbot
3667 followers · 271468 posts · Server schleuss.online

Meta introduces Code Llama, an AI tool aimed at faster coding and debugging - Enlarge (credit: Getty Images | Benj Edwards)

Meta is adding a... - arstechnica.com/?p=1963185

#ai #meta #tech #llama #metaai #llama2 #biz #codellama #aiassistants #textsynthesis #aicodingtools #machinelearning #largelanguagemodels

Last updated 1 year ago

Tech news from Canada · @TechNews
978 followers · 26344 posts · Server mastodon.roitsystems.ca
Sam · @s_mcleod
150 followers · 298 posts · Server aus.social

Code Llama running locally against a git repository with Aider. The open source models just keep getting better.

#llm #llama #codellama #aider #ml #ai

Last updated 1 year ago

Boiling Steam · @boilingsteam
4321 followers · 10348 posts · Server mastodon.cloud
eicker.news #technews · @technews
84 followers · 811 posts · Server eicker.news

»’s Next AI Attack on OpenAI: Meta’s code-generating artificial intelligence model, dubbed , will be and could launch as soon as next week.« theinformation.com/articles/me

#Meta #codellama #opensource #tech #media

Last updated 1 year ago

White House Press Office · @press
75 followers · 336 posts · Server whitehouse.org

I'm so excited to hear the news about ! This incredible new AI model is sure to revolutionize the way developers create programming code. With this open-source code-generating AI model, developers can spend less time coding and more time innovating! techmeme.com/230817/p35#a23081

#codellama #metaplatforms #openai #revolutionizeprogramming #codegenerator

Last updated 1 year ago