Towards Best Practices for Open Datasets for LLM Training
Paper
•
2501.08365
•
Published
•
40
Natural Language Processing with Transformers
AI & ML interests
This organization contains all the models and datasets covered in the book "Natural Language Processing with Transformers".
Recent Activity
View all activity
transformersbook's activity
lvwerra
authored
a
paper
2 days ago
thomwolf
authored
a
paper
2 days ago
Post
3267
I was initially pretty sceptical about Meta's Coconut paper [1] because the largest perf gains were reported on toy linguistic problems. However, these results on machine translation are pretty impressive!
https://x.com/casper_hansen_/status/1875872309996855343
Together with the recent PRIME method [2] for scaling RL, reasoning for open models is looking pretty exciting for 2025!
[1] Training Large Language Models to Reason in a Continuous Latent Space (2412.06769)
[2] https://huggingface.co/blog/ganqu/prime
https://x.com/casper_hansen_/status/1875872309996855343
Together with the recent PRIME method [2] for scaling RL, reasoning for open models is looking pretty exciting for 2025!
[1] Training Large Language Models to Reason in a Continuous Latent Space (2412.06769)
[2] https://huggingface.co/blog/ganqu/prime
Post
2130
This paper (
HuatuoGPT-o1, Towards Medical Complex Reasoning with LLMs (2412.18925)) has a really interesting recipe for inducing o1-like behaviour in Llama models:
* Iteratively sample CoTs from the model, using a mix of different search strategies. This gives you something like Stream of Search via prompting.
* Verify correctness of each CoT using GPT-4o (needed because exact match doesn't work well in medicine where there are lots of aliases)
* Use GPT-4o to reformat the concatenated CoTs into a single stream that includes smooth transitions like "hmm, wait" etc that one sees in o1
* Use the resulting data for SFT & RL
* Use sparse rewards from GPT-4o to guide RL training. They find RL gives an average ~3 point boost across medical benchmarks and SFT on this data already gives a strong improvement.
Applying this strategy to other domains could be quite promising, provided the training data can be formulated with verifiable problems!
* Iteratively sample CoTs from the model, using a mix of different search strategies. This gives you something like Stream of Search via prompting.
* Verify correctness of each CoT using GPT-4o (needed because exact match doesn't work well in medicine where there are lots of aliases)
* Use GPT-4o to reformat the concatenated CoTs into a single stream that includes smooth transitions like "hmm, wait" etc that one sees in o1
* Use the resulting data for SFT & RL
* Use sparse rewards from GPT-4o to guide RL training. They find RL gives an average ~3 point boost across medical benchmarks and SFT on this data already gives a strong improvement.
Applying this strategy to other domains could be quite promising, provided the training data can be formulated with verifiable problems!
- 1 reply
Post
6740
We outperform Llama 70B with Llama 3B on hard math by scaling test-time compute 🔥
How? By combining step-wise reward models with tree search algorithms :)
We show that smol models can match or exceed the performance of their much larger siblings when given enough "time to think"
We're open sourcing the full recipe and sharing a detailed blog post.
In our blog post we cover:
📈 Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time.
🎄 Diverse Verifier Tree Search (DVTS): An unpublished extension we developed to the verifier-guided tree search technique. This simple yet effective method improves diversity and delivers better performance, particularly at large test-time compute budgets.
🧭 Search and Learn: A lightweight toolkit for implementing search strategies with LLMs and built for speed with vLLM
Here's the links:
- Blog post: HuggingFaceH4/blogpost-scaling-test-time-compute
- Code: https://github.com/huggingface/search-and-learn
Enjoy!
How? By combining step-wise reward models with tree search algorithms :)
We show that smol models can match or exceed the performance of their much larger siblings when given enough "time to think"
We're open sourcing the full recipe and sharing a detailed blog post.
In our blog post we cover:
📈 Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time.
🎄 Diverse Verifier Tree Search (DVTS): An unpublished extension we developed to the verifier-guided tree search technique. This simple yet effective method improves diversity and delivers better performance, particularly at large test-time compute budgets.
🧭 Search and Learn: A lightweight toolkit for implementing search strategies with LLMs and built for speed with vLLM
Here's the links:
- Blog post: HuggingFaceH4/blogpost-scaling-test-time-compute
- Code: https://github.com/huggingface/search-and-learn
Enjoy!
- 2 replies
Post
4869
We are proud to announce
HuggingFaceFW/fineweb-2: A sparkling update to
HuggingFaceFW/fineweb with 1000s of 🗣️languages.
We applied the same data-driven approach that led to SOTA English performance in🍷 FineWeb to thousands of languages.
🥂 FineWeb2 has 8TB of compressed text data and outperforms other multilingual datasets in our experiments.
The dataset is released under the permissive 📜 ODC-By 1.0 license, and the 💻 code to reproduce it and our evaluations is public.
We will very soon announce a big community project, and are working on a 📝 blogpost walking you through the entire dataset creation process. Stay tuned!
In the mean time come ask us question on our chat place: HuggingFaceFW/discussion
H/t @guipenedo @hynky @lvwerra as well as @vsabolcec Bettina Messmer @negar-foroutan and @mjaggi
We applied the same data-driven approach that led to SOTA English performance in🍷 FineWeb to thousands of languages.
🥂 FineWeb2 has 8TB of compressed text data and outperforms other multilingual datasets in our experiments.
The dataset is released under the permissive 📜 ODC-By 1.0 license, and the 💻 code to reproduce it and our evaluations is public.
We will very soon announce a big community project, and are working on a 📝 blogpost walking you through the entire dataset creation process. Stay tuned!
In the mean time come ask us question on our chat place: HuggingFaceFW/discussion
H/t @guipenedo @hynky @lvwerra as well as @vsabolcec Bettina Messmer @negar-foroutan and @mjaggi
-
- 2 replies
Post
1263
Exponentially growing number of open-source AI models over the course of the past 30 months – from a few thousands to over 1 million and more
Interactive data viz: huggingface/open-source-ai-year-in-review-2024
Interactive data viz: huggingface/open-source-ai-year-in-review-2024
Post
1440
Most liked and most downloaded open-source AI models from 2022 to 2024
Interactive viz: https://aiworld.eu/embed/model/model/treemap
Discussion: huggingface/open-source-ai-year-in-review-2024
Interactive viz: https://aiworld.eu/embed/model/model/treemap
Discussion: huggingface/open-source-ai-year-in-review-2024
Post
1688
Interesting long read from
@evanmiller-anthropic
on having a better founded statistical approach to Language Model Evaluations:
https://www.anthropic.com/research/statistical-approach-to-model-evals
Worth a read if you're into LLM evaluations!
Cc @clefourrier
https://www.anthropic.com/research/statistical-approach-to-model-evals
Worth a read if you're into LLM evaluations!
Cc @clefourrier
- 1 reply
Post
1438
Very exciting new
mistralai/Pixtral-Large-Instruct-2411 model from Mistral-AI
Impressive performances, huge congrats @patrickvonplaten @sgvaze @pandora-s @devendrachaplot @sophiamyang and team!
Very nice to have SOTA Multilingual OCR and Chart understanding in an open-weights model
Impressive performances, huge congrats @patrickvonplaten @sgvaze @pandora-s @devendrachaplot @sophiamyang and team!
Very nice to have SOTA Multilingual OCR and Chart understanding in an open-weights model
lvwerra
authored
a
paper
3 months ago
Post
4158
Parents in the 1990: Teach the kids to code
Parents now: Teach the kids to fix the code when it starts walking around 🤖✨
Parents now: Teach the kids to fix the code when it starts walking around 🤖✨
- 2 replies
thomwolf
authored
a
paper
7 months ago
lvwerra
authored
a
paper
7 months ago
Post
4572
[New crazy blog post alert] We are releasing an extensive blog post on the science of creating high quality web-scale datasets, detailing all the steps and learnings that came in our recent 15 trillion tokens 🍷FineWeb release
Inspired by the distill.pub interactive graphics papers, we settled to write the most extensive, enjoyable and in-depth tech report we could draft on so prepare for a 45-mmin read with interactive graphics and all.
And it's not all, in this article we also introduce 📚FineWeb-Edu a filtered subset of Common Crawl with 1.3T tokens containing only web pages with very high educational content. Up to our knowledge, FineWeb-Edu out-performs all openly release web-scale datasets by a significant margin on knowledge- and reasoning-intensive benchmarks like MMLU, ARC, and OpenBookQA
We also make a number of surprising observations on the "quality" of the internet it-self which may challenge some of the general assumptions on web data (not saying more, I'll let you draw your conclusions ;)
HuggingFaceFW/blogpost-fineweb-v1
Inspired by the distill.pub interactive graphics papers, we settled to write the most extensive, enjoyable and in-depth tech report we could draft on so prepare for a 45-mmin read with interactive graphics and all.
And it's not all, in this article we also introduce 📚FineWeb-Edu a filtered subset of Common Crawl with 1.3T tokens containing only web pages with very high educational content. Up to our knowledge, FineWeb-Edu out-performs all openly release web-scale datasets by a significant margin on knowledge- and reasoning-intensive benchmarks like MMLU, ARC, and OpenBookQA
We also make a number of surprising observations on the "quality" of the internet it-self which may challenge some of the general assumptions on web data (not saying more, I'll let you draw your conclusions ;)
HuggingFaceFW/blogpost-fineweb-v1
- 1 reply
lvwerra
authored
a
paper
8 months ago
Post
4880
Is is time for the open-source AI robots revolution 🚀?
With @haixuantao and @Leyo we’ve been playing with a low-cost DJI robot controlled by three local open-source AI models (Whisper, Idefics2, Parler-TTS - all Apache2) and orchestrated by Dora-cs.
Links to find all the hardware/software we used in the demo:
- robot control framework – dora-rs: https://github.com/dora-rs/dora
- speech-to-text model – whisper: openai/whisper-base
- vision-text model – Idefics2: HuggingFaceM4/idefics2-8b-AWQ
- text-to-speech model – ParlerTTS mini: parler-tts/parler_tts_mini_v0.1
- robot: https://dji.com/robomaster-s1
- code gist: https://gist.github.com/haixuanTao/860e1740245dc2c8dd85b496150a9320
- Larger codebase: dora-rs/dora-idefics2
- laptop/pc: any with a recent GPU card (our has a RTX 4090)
Enjoy!
With @haixuantao and @Leyo we’ve been playing with a low-cost DJI robot controlled by three local open-source AI models (Whisper, Idefics2, Parler-TTS - all Apache2) and orchestrated by Dora-cs.
Links to find all the hardware/software we used in the demo:
- robot control framework – dora-rs: https://github.com/dora-rs/dora
- speech-to-text model – whisper: openai/whisper-base
- vision-text model – Idefics2: HuggingFaceM4/idefics2-8b-AWQ
- text-to-speech model – ParlerTTS mini: parler-tts/parler_tts_mini_v0.1
- robot: https://dji.com/robomaster-s1
- code gist: https://gist.github.com/haixuanTao/860e1740245dc2c8dd85b496150a9320
- Larger codebase: dora-rs/dora-idefics2
- laptop/pc: any with a recent GPU card (our has a RTX 4090)
Enjoy!
- 4 replies
Post
5059
Introducing Zephyr 141B-A35B 🪁:
HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1
Yesterday, Mistral released their latest base model (via magnet link of course 😅) and the community quickly converted it to transformers format and pushed it to the Hub: mistral-community/Mixtral-8x22B-v0.1
Early evals of this model looked extremely strong, so we teamed up with Argilla and KAIST AI to cook up a Zephyr recipe with a few new alignment techniques that came out recently:
🧑🍳 Align the base model with Odds Ratio Preference Optimisation (ORPO). This novel algorithm developed by @JW17 and @nlee-208 and @j6mes and does not require an SFT step to achieve high performance and is thus much more computationally efficient than methods like DPO and PPO.
🦫 Use a brand new dataset of 7k high-quality, multi-turn preferences that has been developed by our friends at Argilla. To create this dataset, they took the excellent Capybara SFT dataset from @LDJnr LDJnr/Capybara and converted it into a preference dataset by augmenting the final turn with responses from new LLMs that were then ranked by GPT-4.
What we find especially neat about this approach is that training on 7k samples only takes ~1.3h on 4 H100 nodes, yet produces a model that is very strong on chat benchmarks like IFEval and BBH.
Kudos to @alvarobartt @JW17 and @nlee-208 for this very nice and fast-paced collab!
For more details on the paper and dataset, checkout our collection: HuggingFaceH4/zephyr-orpo-6617eba2c5c0e2cc3c151524
HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1
Yesterday, Mistral released their latest base model (via magnet link of course 😅) and the community quickly converted it to transformers format and pushed it to the Hub: mistral-community/Mixtral-8x22B-v0.1
Early evals of this model looked extremely strong, so we teamed up with Argilla and KAIST AI to cook up a Zephyr recipe with a few new alignment techniques that came out recently:
🧑🍳 Align the base model with Odds Ratio Preference Optimisation (ORPO). This novel algorithm developed by @JW17 and @nlee-208 and @j6mes and does not require an SFT step to achieve high performance and is thus much more computationally efficient than methods like DPO and PPO.
🦫 Use a brand new dataset of 7k high-quality, multi-turn preferences that has been developed by our friends at Argilla. To create this dataset, they took the excellent Capybara SFT dataset from @LDJnr LDJnr/Capybara and converted it into a preference dataset by augmenting the final turn with responses from new LLMs that were then ranked by GPT-4.
What we find especially neat about this approach is that training on 7k samples only takes ~1.3h on 4 H100 nodes, yet produces a model that is very strong on chat benchmarks like IFEval and BBH.
Kudos to @alvarobartt @JW17 and @nlee-208 for this very nice and fast-paced collab!
For more details on the paper and dataset, checkout our collection: HuggingFaceH4/zephyr-orpo-6617eba2c5c0e2cc3c151524
Post
3126
Very interesting model just released by MyShell:
jetmoe/jetmoe-8b . It's a 8B-parameters MoE LLM so 2.2B active parameters, really efficient.
Main characteristics:
- impressive performances for its size (beating meta-llama/Llama-2-7b and huggyllama/llama-13b)
- combine Mixture of Attention heads (MoA) and Mixture of MLP Experts (MoE) – 8 experts with 2 being active for each token
- trained on a rather limited 1.25T tokens from publicly available datasets – training recipe follows the MiniCPM's two-phases training method => first time I see this for a 2B+ model
- $100k to train
- open weights - open sharing of recipes - open dataset - open code => ♡
- still interesting room to improve performances (be it only by training longer)
Links:
- report: https://research.myshell.ai/jetmoe
- model: jetmoe/jetmoe-8b
- code: https://github.com/myshell-ai/JetMoE
Note: I actually detailed all of the MiniCPM schedule, Mixture-of-expert (MoE) and many of the datasets used in this work in my recent little guide to building LLMs in 2024, so feel free to check it out if you want to learn more on these topics: https://www.youtube.com/watch?v=2-SPH9hIKT8
Main characteristics:
- impressive performances for its size (beating meta-llama/Llama-2-7b and huggyllama/llama-13b)
- combine Mixture of Attention heads (MoA) and Mixture of MLP Experts (MoE) – 8 experts with 2 being active for each token
- trained on a rather limited 1.25T tokens from publicly available datasets – training recipe follows the MiniCPM's two-phases training method => first time I see this for a 2B+ model
- $100k to train
- open weights - open sharing of recipes - open dataset - open code => ♡
- still interesting room to improve performances (be it only by training longer)
Links:
- report: https://research.myshell.ai/jetmoe
- model: jetmoe/jetmoe-8b
- code: https://github.com/myshell-ai/JetMoE
Note: I actually detailed all of the MiniCPM schedule, Mixture-of-expert (MoE) and many of the datasets used in this work in my recent little guide to building LLMs in 2024, so feel free to check it out if you want to learn more on these topics: https://www.youtube.com/watch?v=2-SPH9hIKT8
- 1 reply
Post
1996
Little know gem: the Open-source Cookbook
A collection of notebooks for building practical AI applications using open-source tools and models: https://lnkd.in/e6m6Jmwu
Doc: https://lnkd.in/e3FE6TUq
Currently contains 16 notebooks in English (and some in Chinese):
1. Using LLM-as-a-judge 🧑⚖️ for an automated and versatile evaluation
2. Create a legal preference dataset
3. Suggestions for Data Annotation with SetFit in Zero-shot Text Classification
4. Implementing semantic cache to improve a RAG system
5. Building A RAG Ebook “Librarian” Using LlamaIndex
6. Stable Diffusion Interpolation
7. Building A RAG System with Gemma, MongoDB and Open Source Models
8. Prompt Tuning with PEFT Library
9. Migrating from OpenAI to Open LLMs Using TGI’s Messages API
10. Automatic Embeddings with TEI through Inference Endpoints
11. Simple RAG for GitHub issues using Hugging Face Zephyr and LangChain
12. Embedding multimodal data for similarity search using 🤗 transformers, 🤗 datasets and FAISS
13. Fine-tuning a Code LLM on Custom Code on a single GPU
14. RAG Evaluation Using Synthetic data and LLM-As-A-Judge
15. Advanced RAG on HuggingFace documentation using LangChain
16. Detecting Issues in a Text Dataset with Cleanlab
A collection of notebooks for building practical AI applications using open-source tools and models: https://lnkd.in/e6m6Jmwu
Doc: https://lnkd.in/e3FE6TUq
Currently contains 16 notebooks in English (and some in Chinese):
1. Using LLM-as-a-judge 🧑⚖️ for an automated and versatile evaluation
2. Create a legal preference dataset
3. Suggestions for Data Annotation with SetFit in Zero-shot Text Classification
4. Implementing semantic cache to improve a RAG system
5. Building A RAG Ebook “Librarian” Using LlamaIndex
6. Stable Diffusion Interpolation
7. Building A RAG System with Gemma, MongoDB and Open Source Models
8. Prompt Tuning with PEFT Library
9. Migrating from OpenAI to Open LLMs Using TGI’s Messages API
10. Automatic Embeddings with TEI through Inference Endpoints
11. Simple RAG for GitHub issues using Hugging Face Zephyr and LangChain
12. Embedding multimodal data for similarity search using 🤗 transformers, 🤗 datasets and FAISS
13. Fine-tuning a Code LLM on Custom Code on a single GPU
14. RAG Evaluation Using Synthetic data and LLM-As-A-Judge
15. Advanced RAG on HuggingFace documentation using LangChain
16. Detecting Issues in a Text Dataset with Cleanlab