💫 Generate RAG data with the Synthetic Data Generator to improve your RAG system! 1️⃣ Generate from your documents, dataset, or dataset description. 2️⃣ Configure it. 3️⃣ Generate the synthetic dataset. 4️⃣ Fine-tune the retrieval and reranking models. 5️⃣ Build a RAG pipeline.
🚀 Argilla v2.6.0 is here! 🎉 Let me show you how EASY it is to export your annotated datasets from Argilla to the Hugging Face Hub. 🤩 Take a look to this quick demo 👇 💁♂️ More info about the release at github.com/argilla-io/a... #AI #MachineLearning #OpenSource #DataScience #HuggingFace #Argilla
🙅♀️ No-code end-to-end example to train your model 1️⃣ Use the Synthetic Data Generator to create your custom dataset 2️⃣ Use AutoTrain to use the generated dataset and train your model Check it here: huggingface.co/blog/synthet...
Where do I get quality data from? We often need to fine-tune models for very specific scenarios. And that’s where the Synthetic Data Generator comes in! Want to see how it works? Watch this quick video (www.youtube.com/watch?v=nXjV...) and get started here: t.co/hJ1b2TsMq0
Synthetic Data Generator - Build Datasets Using Natural Language
YouTube video by Argilla
youtube.com
Pouco a pouco avanzamos! 🚀 Anímovos a contribuir, tan só tedes que entrar na ligazón, ler as instrucións e comezar a anotar ✍ data-is-better-together-fineweb-c.hf.space/share-your-p...
glg - galego - Galician
Join and contribute to the dataset glg - galego - Galician
data-is-better-together-fineweb-c.hf.space
Spanish, Filipino, Amharic, French, German, Basque, Catalan, Galician, Guarani, Telugu, Italian, Pashto, Romanian, Tamil, Urdu, Danish... and many more! All included in the FineWeb2 Community Annotation Sprint! 🔥 💫 Join to build an impactful dataset for your language!
Open Image Preferences released! 🚀 - Open-source dataset for text2image - 10K samples manually evaluated by the HF community. - Binarized format for SFT, DPO, or ORPO. It comes with a nice blog post explaining the steps to pre-process and generate the data, along with the results.
Language is power! A multilingual annotation sprint for hundreds of languages is starting soon! Step up as a Language Lead and help drive this effort for your language. If there's already a Language Lead, stay tuned! Is this the start of a nice community? docs.google.com/forms/d/e/1F...
Language Lead sign-up
At Hugging Face 🤗, we're launching a big community initiative to improve LLM training for many languages. We're looking for Language Leads to help us cultivate specific languages during this initiativ...
docs.google.com
Want to improve your model quality? Implement the data annotation stage in your MLOps effortlessly thanks to the enhanced integration of Argilla with ZenML. ✨Use the latest Argilla features ✨Improve human-in-the-loop workflows ✨Manage datasets, track progress, and coordinate your annotation team
🚀 QwQ-32B-Preview is available on the Hub! > The results are very promising, beating o1-mini. > However, they also have several limitations you might notice even in the demo (I found endless reasoning trying to find out the number of 'r' in 🍓). So, let's see how they deal with them.
It's pretty sad to see the negative sentiment towards Hugging Face on this platform due to a dataset put by one of the employees. I want to write a small piece. 🧵 Hugging Face empowers everyone to use AI to create value and is against monopolization of AI it's a hosting platform above all.
Hugging Face inference endpoints now support CPU deployment for llama.cpp 🚀 🚀 Why this is a huge deal? Llama.cpp is well-known for running very well on CPU. If you're running small models like Llama 1B or embedding models, this will definitely save tons of money 💰 💰
🎨 Help to build an image preference dataset! > Goal: Release an open-source image dataset, enabling the entire community to benefit from it. > Requirements: All you need is a Hugging Face account and a willingness to contribute. More in 🧵
Let's make AI more inclusive. At @huggingface.bsky.social we'll launch a huge community sprint soon to build high-quality training datasets for many languages. We're looking for Language Leads to help with outreach. Find your language and nominate yourself: forms.gle/iAJVauUQ3FN8...
"Naftali was assigned to train AI to recognize and weed out pornography, hate speech and excessive violence, which meant sifting through the worst of the worst content online for hours on end." So much of AI is based on exploiting workers in precarious conditions 😔 www.cbsnews.com/news/labeler...
Labelers training AI say they're overworked, underpaid and exploited by big American tech companies
Digital workers in Kenya had to sift through horrific online content to train AI, but say they were underpaid, overworked, and got inadequate mental health support. So they're fighting back.
cbsnews.com
Argilla has reached the 4K stars on GitHub! ✨ Thanks to everyone for the support! We'll continue shipping new updates for you to curate your data easily ✍️ And, for sure, your feedback is more than welcome 🙌
Making SmolLM2 more reproducible: open-sourcing our training & evaluation toolkit 🛠️ github.com/huggingface/... Pre-training & evaluation code, synthetic data generation pipelines, post-training scripts, on-device tools & demos Apache 2.0. V2 data mix coming soon! Which tools should we add next?
GitHub - huggingface/smollm: Everything about the SmolLM & SmolLM2 family of models
Everything about the SmolLM & SmolLM2 family of models - GitHub - huggingface/smollm: Everything about the SmolLM & SmolLM2 family of models
github.com
Go to your HF profile and spot the differences. Could you find them? . . . Yes, you're right! You can now add your Bluesky account and also check your recent activity 🙌
We're big on transparency and love all things open at @huggingface.bsky.social so we thought, why not share our internal mission, vision, and values with everyone? Take a read and let us know what you think! It's such a special place!
A cool space from the community to explore the many open datasets we released from building Tulu 3 (SOTA open post training recipe).
Daniel Vila@dvilasuero.hf.co · 2y agoTülu 3, my favourite Open Source AI release ever As @natolambert.bsky.social says, it sets the next era in open post-training. My highlight? the data generation & open datasets Want to deep dive into the data? Here's an Argilla @huggingface.bsky.social Space huggingface.co/spaces/argil...
The SmolLM2 recipe isn't more of a secret! SmolTalk has been released on the Hub. It uses a mix of public and synthetic data, including Magpie Ultra using distilabel 🚀 huggingface.co/datasets/Hug...
HuggingFaceTB/smoltalk · Datasets at Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
Excited to announce the SFT dataset used for @huggingface.bsky.social SmolLM2! The dataset for SmolLM2 was created by combining multiple existing datasets and generating new synthetic datasets, including MagPie Ultra v1.0, using distilabel. Check out the dataset: huggingface.co/datasets/Hug...
HuggingFaceTB/smoltalk · Datasets at Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
✨ Time to introduce myself to the new followers! I'm working on ML and advocacy. Till now, I have focused on building Argilla and distilabel. But I'll also share updates and cool stuff from the AI community, tools, or notebooks. P.S. Maybe a bit of my 🐕 and 🎮 too!
synthetic data on easy_mode We noticed that folk were building synthetic datasets in a common way. Basically going from prompt to dataset, and iterating. So we implemented a friendly abstraction to help you learn and improve this flow.
You might not know me yet, and that’s perfect—no bias here ✨. So, I just have a question for you. > What three key aspects would make an annotation tool ideal for you? Feel free to share your thoughts! I'm reading you 🤗
