Burla รีทวีตแล้ว
@sundeep court says it's fair use? cool. i just built a 4K cpu cluster using @Burla_dev to screenshot academia into dust.
English
Burla
39 posts
Burla
@Burla_dev
Scale Python to 10,000 VM's with one line of code: https://t.co/gEyAjO2qgc
Boston เข้าร่วม Ağustos 2023
40 กำลังติดตาม15 ผู้ติดตาม
@TheTranscript_ the accelerated demand for compute will really start when people are running 10 thousands parallel AI agents using @Burla_dev
English
$NVDA CEO: "AI inference token generation has surged tenfold in just one year, and as AI agents become mainstream, the demand for AI computing will accelerate."
English
間もなくサービスを開始されるという連絡を頂きました。
burla.dev
thanks to Jake for ping @Burla_dev github.com/Burla-Cloud/bu…
日本語
@singh__medhavi Burla is the world's simplest cluster compute software. With zero setup, and just one line of code, Burla scales your python programs to thousands of computers in the cloud.
English
@imakash_skyy Burla is the world's simplest cluster compute software. With zero setup, and just one line of code, Burla scales your python programs to thousands of computers in the cloud.
English
Drop your startup idea or dm . Will invest if it's worth investing.
Will work for it.
It can be at any stage or industry.
#startup #investment #seedfund #funding #business
English
@siyabuilt Burla is the world's simplest cluster compute software.
Scale across thousands of computers in the cloud,
with zero setup, and just one line of code.
English
@omarsar0 I know a cloud abstraction that could help docs.burla.dev/Example:%20Mas…
English
Batch Prompting: Efficient Inference with LLM APIs
Batch prompting helps to reduce the inference token and time costs while achieving better or comparable performance. Love to find these neat little tricks on efficiency gains during inference with LLMs.
arxiv.org/abs/2301.08721
English
English
@WenhuChen If your students need any help scaling their code, just let me know! Here in Boston, I've been seeing tons of PhDs and researchers training on only a small fraction of their data because of campus resource competition. Thought I'd mention in case it's helpful for your situation!
English
Working on LLMs is really stressing. I have seen my research ideas being scooped almost on a weekly basis 🥲.
English
Burla รีทวีตแล้ว
New Startup Listed 🚀
Burla
Cluster compute software for python beginners
Vote for them below 👇
startupstage.app/startups/burla
English
Monte Carlo simulations:
A way to estimate portfolio values over 1000s of price paths.
Yesterday I published a step-by-step walk through of using Python to do it.
You can read it free:
English
@pyquantnews Are there any workloads you have that would greatly benefit from large levels of parallelism? Not a few cores but thousands of workers? I'm open-sourcing Burla in the next few weeks and want to build out some useful tutorials :)
English
@Burla_dev I have but these tend to run fast on modern hardware
English
@yacineMTB If y'all need to do batch inference on hundreds of GPUs in parallel take a look at this tutorial. OS version coming out in the next couple of weeks.
docs.burla.dev/Example:%20Mas…
English
@chamath If you're seeking a serverless function for running large-scale LLM inference, give Burla a try. Here is a tutorial on massively parallelizing inference (200 GPUs) using Mistral-7B.
docs.burla.dev/Example:%20Mas…
English
AI clouds will be a clear winner in this cycle. Whoever can deliver the cheapest tokens/second will build a huge company.
sunny madra@sundeep
Inference is the new currency...
English
@sundeep Lowering the entry barrier for running large machines in the cloud is crucial. Currently, we are concentrating on refining the most straightforward cloud abstraction. Any Python developer, irrespective of their skill set, should be able to effectively leverage cloud resources.
English
Inference is the new currency...
ZenCap@kaizen_cap
$MSFT print pretty bullish for SaaS. AI application layer way bigger than people thought if the 6% contribution is mostly inference. Net new workload growth going to benefit all the infra players. The line between non-AI and AI workloads is becoming fungible.
English