Savin

New paper: 🧭 Introducing VAL-Bench: Measuring Value Alignment in Language Models. A benchmark that measures the consistency in language model expression of human values when prompted to justify opposing positions on real-life issues. Work with @amang0112358 and Denny O'Shea! (1/7)

@ludwigABAP @semiozz Would love to know the missing pieces!

@semiozz nah... no real final decision yet but it seems we'll be using metaflow + our own stuff for all that we'd miss, since we already use metaflow in some other projects

In the "setting up an inference server for 300 concurrent internal users with limited on-prem hardware where if we don't optimize the token throughput and context window length, my first child will be executed" mines

What powers Netflix-level ML/AI workloads? Dive into Outerbounds’ architecture and discover why platform engineers are in love with it: outerbounds.com/blog/architect…

HIPAA hooray! Outerbounds is now HIPAA certified! 🎉 Build, scale, and deploy ML/AI solutions in healthcare with full confidence in your data privacy and compliance: hubs.ly/Q02TSNP40

🚀 Exciting updates from the @PyTorch Conference in San Francisco, joined by Eddie, Ville, and Omed! Eddie Mattia shares key highlights on our latest features launched this week, including [@]checkpoint for robust training jobs and [@]torchrun for distributed training, addressing the critical need for fault tolerance as model sizes scale. GPUs will fail, but our new checkpoint functionality—available in open-source Metaflow and enhanced on the Outerbounds platform—helps ensure deep learning jobs can resume seamlessly after interruptions. We’ve received great feedback from the community, and we're excited to see what people will build with these tools! Read more from our Wednesday announcement blog: outerbounds.com/blog/scale-ml-…

🚀🚀🚀

💫How have GenAI and foundation models changed the practice of ML engineering? Come find out from @chipro in the @OuterboundsHQ July fireside chat! 🔥 lu.ma/a3ynbto1?utm_s…

@spimescape anytime! also community support is just a slack channel away at chat.metaflow.org!

@SavinGoyal Oh wow thank you- I did not know this! Will take a look esp since we have kind of built our whole ML infra ecosystem around Cloud Composer/GCPs manages AirFlow soln.

Has anyone migrated from Metaflow to AirFlow or vice versa?

@k_urushi7 How can I help!

@s_tat1204 That’s true! Even requests is only a soft dependency to power the UI. You can run metaflow locally without installing any of its dependencies!

metaflow、各種インフラ（AWS/GCS/...）等に対するplugin抜くと、requestsとboto3にしか依存してなさそう。すごいな。 boto3も実はAWSを使うときしか参照されてなさそうなので、実質requestsのみにしか依存していないのかも。 3rd party moduleへの依存をしないという強い意志を感じる。 #L52" target="_blank" rel="nofollow noopener">github.com/Netflix/metafl…

@ayushunleashed 🙋‍♂️

anyone who uses metaflow here?? just want to connect with datascience nerds to ask questions or brainstorm some ideas which could be solved with nlp.

📢✨Metaflow, coming to a city near you. Check out our spring calendar and join us in SF, Austin, Boston, London, Paris, Berlin, or Amsterdam! 🤗 outerbounds.com/blog/ob-on-the…

I was asked by @hugobowne to "steel-man" the argument around AI x-risk -- here's my attempt at (briefly and simply) doing that: youtube.com/live/6LXw2bepr…

OK, let's start getting real about LLMs again. Join me and @jeremyphoward later this week as we discuss the current and future states of generative AI. lu.ma/byxyzfrr?utm_s… We'll talk about * What on earth we’ve just witnessed at OpenAI, why it’s important (and why it’s not!), * Cost-benefit analyses of using closed models (such as Chat-GPT and Claude) and OSS models (such as Llama 2 and Mistral), * Key concerns around regulatory capture and the future of OSS LLMs, * What LLM tools and techniques you need to know to future-proof yourself in this rapidly changing landscape, * How AI education will need to change to make LLMs actually uncool so that many people can actually use them! And much more!

Who else is at @exploratorium to see @vtuulos become a sofa salesman?

Faced with capacity issues at AWS, the Outerbounds team leveraged CoreWeave A100's to train a Dolly foundational model with Metaflow. Learn more about the history of open-source models and LLM fine-tuning in this article from our friends at Outerbounds: hubs.la/Q01V7rkK0

Unsere #MachineLearning (ML)-Ingenieur:innen und Datenwissenschaftler:innen haben kürzlich ein neues ML-Modell entwickelt. Welche Rolle hierbei #Metaflow spielt und vor welchen Herausforderungen sie standen, erzählen sie in unserem Tech-Blog. ➡️ pro7.de/35a

A major release, Metaflow 2.9, is out today 🎉 New: Build large, advanced ML systems by connecting flows and make them react to events in real-time! 🛰️ outerbounds.com/blog/metaflow-…

As developers continue building LM-driven apps, companies will adapt, not replace, their existing ML and technology stacks. As such, tools that support both LM and classical ML use cases may be ideal. More great ideas from @OuterboundsHQ here: outerbounds.com/blog/llm-infra…