Reposted by Manos Koukoumidis
So excited to announce the DCVLR (Data Curation for Vision-Language Reasoning) competition at #NeurIPS2025, led by @oumi-pbc.bsky.social and Lambda AI!
🌟open-data 🌟
🤖 open-models 🤖
💻 open-source 💻
💪anyone can compete for free 💪
dcvlr-neurips.github.io
🧵 1 / n
🌟open-data 🌟
🤖 open-models 🤖
💻 open-source 💻
💪anyone can compete for free 💪
dcvlr-neurips.github.io
🧵 1 / n
DCVLR: Data Curation for Vision Language Reasoning - NeurIPS 2025 Competition
Join the DCVLR NeurIPS 2025 Competition. Advance visual reasoning in VLMs through data curation.
dcvlr-neurips.github.io
June 18, 2025 at 2:18 PM
So excited to announce the DCVLR (Data Curation for Vision-Language Reasoning) competition at #NeurIPS2025, led by @oumi-pbc.bsky.social and Lambda AI!
🌟open-data 🌟
🤖 open-models 🤖
💻 open-source 💻
💪anyone can compete for free 💪
dcvlr-neurips.github.io
🧵 1 / n
🌟open-data 🌟
🤖 open-models 🤖
💻 open-source 💻
💪anyone can compete for free 💪
dcvlr-neurips.github.io
🧵 1 / n
Reposted by Manos Koukoumidis
🚀 Custom Evaluations Made Easy with Oumi 🚀
Build custom evaluations for any model in <50 lines of code! 🙌
✅ Simple config change to evaluate models like GPT-4o, Claude 3.7, LLaMA 405B
✅ Learn how we evaluated SOTA LLMs as hallucination classifiers🧠✨
👉 Check it out: oumi.ai/blog/posts/c...
Build custom evaluations for any model in <50 lines of code! 🙌
✅ Simple config change to evaluate models like GPT-4o, Claude 3.7, LLaMA 405B
✅ Learn how we evaluated SOTA LLMs as hallucination classifiers🧠✨
👉 Check it out: oumi.ai/blog/posts/c...
Oumi - Build Custom Evaluations for any Open or Closed Model in just 50 Lines of Code
Use Oumi to build custom evaluations for any open or closed model, in just 50 lines of code
oumi.ai
March 25, 2025 at 9:56 PM
🚀 Custom Evaluations Made Easy with Oumi 🚀
Build custom evaluations for any model in <50 lines of code! 🙌
✅ Simple config change to evaluate models like GPT-4o, Claude 3.7, LLaMA 405B
✅ Learn how we evaluated SOTA LLMs as hallucination classifiers🧠✨
👉 Check it out: oumi.ai/blog/posts/c...
Build custom evaluations for any model in <50 lines of code! 🙌
✅ Simple config change to evaluate models like GPT-4o, Claude 3.7, LLaMA 405B
✅ Learn how we evaluated SOTA LLMs as hallucination classifiers🧠✨
👉 Check it out: oumi.ai/blog/posts/c...
Reposted by Manos Koukoumidis
Looking for simple, robust model evaluations? Look no further!
📊Standardize benchmarks
🗣️Generative evals
🔧Customizable evals
Oumi has them all!
Today in our Weekly Walkthrough, @taenin.bsky.social will give you an overview of Oumi’s evaluation framework.
🎥 Watch now: youtu.be/GhHmtjMw-l4
📊Standardize benchmarks
🗣️Generative evals
🔧Customizable evals
Oumi has them all!
Today in our Weekly Walkthrough, @taenin.bsky.social will give you an overview of Oumi’s evaluation framework.
🎥 Watch now: youtu.be/GhHmtjMw-l4
Evaluate Models with Oumi [walkthrough]
YouTube video by Oumi
youtu.be
March 19, 2025 at 5:23 PM
Looking for simple, robust model evaluations? Look no further!
📊Standardize benchmarks
🗣️Generative evals
🔧Customizable evals
Oumi has them all!
Today in our Weekly Walkthrough, @taenin.bsky.social will give you an overview of Oumi’s evaluation framework.
🎥 Watch now: youtu.be/GhHmtjMw-l4
📊Standardize benchmarks
🗣️Generative evals
🔧Customizable evals
Oumi has them all!
Today in our Weekly Walkthrough, @taenin.bsky.social will give you an overview of Oumi’s evaluation framework.
🎥 Watch now: youtu.be/GhHmtjMw-l4
Reposted by Manos Koukoumidis
Need to label a dataset? Looking to run inference at scale? You’re in luck! 🍀
@taenin.bsky.social tackles Batch Prediction in this edition of our Weekly Walkthrough, showing how you can quickly run inference over your data within the Oumi platform!
🎥 Watch now: youtu.be/3Yg3ycxCEYQ
@taenin.bsky.social tackles Batch Prediction in this edition of our Weekly Walkthrough, showing how you can quickly run inference over your data within the Oumi platform!
🎥 Watch now: youtu.be/3Yg3ycxCEYQ
Let's run batch prediction with Oumi [walkthrough]
YouTube video by Oumi
youtu.be
March 12, 2025 at 4:14 PM
Need to label a dataset? Looking to run inference at scale? You’re in luck! 🍀
@taenin.bsky.social tackles Batch Prediction in this edition of our Weekly Walkthrough, showing how you can quickly run inference over your data within the Oumi platform!
🎥 Watch now: youtu.be/3Yg3ycxCEYQ
@taenin.bsky.social tackles Batch Prediction in this edition of our Weekly Walkthrough, showing how you can quickly run inference over your data within the Oumi platform!
🎥 Watch now: youtu.be/3Yg3ycxCEYQ
Reposted by Manos Koukoumidis
BOS = Beginning of sequence/sentence
Also whoever decided to make Phi have 10 KV heads instead of a multiple of 4 or 8, why do you hate multiple GPUs 😭
Also whoever decided to make Phi have 10 KV heads instead of a multiple of 4 or 8, why do you hate multiple GPUs 😭
March 12, 2025 at 9:44 PM
BOS = Beginning of sequence/sentence
Also whoever decided to make Phi have 10 KV heads instead of a multiple of 4 or 8, why do you hate multiple GPUs 😭
Also whoever decided to make Phi have 10 KV heads instead of a multiple of 4 or 8, why do you hate multiple GPUs 😭
Reposted by Manos Koukoumidis
Apparently a lot of you were dreaming about DeepSeek's R1 model 👀
@jgreer-oumi.bsky.social's MiniMath-R1-1.5B has over 750 downloads since we released it last week!
Have you used it to train your own DeepSeek R1 model yet?
@jgreer-oumi.bsky.social's MiniMath-R1-1.5B has over 750 downloads since we released it last week!
Have you used it to train your own DeepSeek R1 model yet?
Dreaming about DeepSeek R1 this month? ☁️
We collected 650,000+ R1 prompts and responses for you to train your own DeepSeek-R1 model.
With this data we made MiniMath-R1-1.5B, the top MMLU-Pro-Math model at <=1.5B parameters, achieving an accuracy of 44.4%!
Data, model, and notebook in thread! 🙌
We collected 650,000+ R1 prompts and responses for you to train your own DeepSeek-R1 model.
With this data we made MiniMath-R1-1.5B, the top MMLU-Pro-Math model at <=1.5B parameters, achieving an accuracy of 44.4%!
Data, model, and notebook in thread! 🙌
March 4, 2025 at 5:39 PM
Apparently a lot of you were dreaming about DeepSeek's R1 model 👀
@jgreer-oumi.bsky.social's MiniMath-R1-1.5B has over 750 downloads since we released it last week!
Have you used it to train your own DeepSeek R1 model yet?
@jgreer-oumi.bsky.social's MiniMath-R1-1.5B has over 750 downloads since we released it last week!
Have you used it to train your own DeepSeek R1 model yet?
Reposted by Manos Koukoumidis
Dreaming about DeepSeek R1 this month? ☁️
We collected 650,000+ R1 prompts and responses for you to train your own DeepSeek-R1 model.
With this data we made MiniMath-R1-1.5B, the top MMLU-Pro-Math model at <=1.5B parameters, achieving an accuracy of 44.4%!
Data, model, and notebook in thread! 🙌
We collected 650,000+ R1 prompts and responses for you to train your own DeepSeek-R1 model.
With this data we made MiniMath-R1-1.5B, the top MMLU-Pro-Math model at <=1.5B parameters, achieving an accuracy of 44.4%!
Data, model, and notebook in thread! 🙌
February 25, 2025 at 5:01 PM
Dreaming about DeepSeek R1 this month? ☁️
We collected 650,000+ R1 prompts and responses for you to train your own DeepSeek-R1 model.
With this data we made MiniMath-R1-1.5B, the top MMLU-Pro-Math model at <=1.5B parameters, achieving an accuracy of 44.4%!
Data, model, and notebook in thread! 🙌
We collected 650,000+ R1 prompts and responses for you to train your own DeepSeek-R1 model.
With this data we made MiniMath-R1-1.5B, the top MMLU-Pro-Math model at <=1.5B parameters, achieving an accuracy of 44.4%!
Data, model, and notebook in thread! 🙌
I’m blown away by the response! 🤯 Within 24 hours, @oumi-pbc.bsky.social became the top trending repository on GitHub! ⭐
The community is ready for 100% truly open AI. Let’s keep the momentum going. Together, we can build the open-source AI the world truly needs. 🌍 🚀
The community is ready for 100% truly open AI. Let’s keep the momentum going. Together, we can build the open-source AI the world truly needs. 🌍 🚀
February 3, 2025 at 10:29 PM
I’m blown away by the response! 🤯 Within 24 hours, @oumi-pbc.bsky.social became the top trending repository on GitHub! ⭐
The community is ready for 100% truly open AI. Let’s keep the momentum going. Together, we can build the open-source AI the world truly needs. 🌍 🚀
The community is ready for 100% truly open AI. Let’s keep the momentum going. Together, we can build the open-source AI the world truly needs. 🌍 🚀
If AI isn’t truly open, it will fail us. We can’t close in a black box our greatest invention yet just so that a few can freely monetize. AI needs its Linux moment, and so we started working towards it. This can only succeed if we all work together!
#oumi #opensource #collaboration
#oumi #opensource #collaboration
January 29, 2025 at 5:07 PM
If AI isn’t truly open, it will fail us. We can’t close in a black box our greatest invention yet just so that a few can freely monetize. AI needs its Linux moment, and so we started working towards it. This can only succeed if we all work together!
#oumi #opensource #collaboration
#oumi #opensource #collaboration