SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model
Paper
β’
2502.02737
β’
Published
β’
110
BigCode Data
non-profit
AI & ML interests
None defined yet.
Recent Activity
View all activity
bigcode-data's activity
lewtunΒ
authored
a
paper
about 19 hours ago
loubnabnlΒ
authored
a
paper
about 19 hours ago
anton-lΒ
authored
a
paper
about 19 hours ago
thomwolfΒ
authored
a
paper
about 19 hours ago
lvwerraΒ
authored
a
paper
about 23 hours ago
Post
9907
We are reproducing the full DeepSeek R1 data and training pipeline so everybody can use their recipe. Instead of doing it in secret we can do it together in the open!
π§ͺ Step 1: replicate the R1-Distill models by distilling a high-quality reasoning corpus from DeepSeek-R1.
π§ Step 2: replicate the pure RL pipeline that DeepSeek used to create R1-Zero. This will involve curating new, large-scale datasets for math, reasoning, and code.
π₯ Step 3: show we can go from base model -> SFT -> RL via multi-stage training.
Follow along: https://github.com/huggingface/open-r1
π§ͺ Step 1: replicate the R1-Distill models by distilling a high-quality reasoning corpus from DeepSeek-R1.
π§ Step 2: replicate the pure RL pipeline that DeepSeek used to create R1-Zero. This will involve curating new, large-scale datasets for math, reasoning, and code.
π₯ Step 3: show we can go from base model -> SFT -> RL via multi-stage training.
Follow along: https://github.com/huggingface/open-r1
- 5 replies
lvwerraΒ
authored
a
paper
22 days ago
thomwolfΒ
authored
a
paper
22 days ago
Post
2188
π€π€ π» Speaking of AI agents ...
...Is easier with the right words ;)
My colleagues @meg @evijit @sasha and @giadap just published a wonderful blog post outlining some of the main relevant notions with their signature blend of value-informed and risk-benefits contrasting approach. Go have a read!
https://huggingface.co/blog/ethics-soc-7
...Is easier with the right words ;)
My colleagues @meg @evijit @sasha and @giadap just published a wonderful blog post outlining some of the main relevant notions with their signature blend of value-informed and risk-benefits contrasting approach. Go have a read!
https://huggingface.co/blog/ethics-soc-7
Post
3809
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
2268
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
joanrodaiΒ
authored
a
paper
about 2 months ago
Post
2385
Introducing ππ
π’π§πππππ‘: the best public math pre-training dataset with 50B+ tokens!
HuggingFaceTB/finemath
Math remains challenging for LLMs and by training on FineMath we see considerable gains over other math datasets, especially on GSM8K and MATH.
We build the dataset by:
π οΈ carefully extracting math data from Common Crawl;
π iteratively filtering and recalling high quality math pages using a classifier trained on synthetic annotations to identify math reasoning and deduction.
We conducted a series of ablations comparing the performance of Llama-3.2-3B-Base after continued pre-training on FineMath and observe notable gains compared to the baseline model and other public math datasets.
We hope this helps advance the performance of LLMs on math and reasoning! π
Weβre also releasing all the ablation models as well as the evaluation code.
HuggingFaceTB/finemath-6763fb8f71b6439b653482c2
HuggingFaceTB/finemath
Math remains challenging for LLMs and by training on FineMath we see considerable gains over other math datasets, especially on GSM8K and MATH.
We build the dataset by:
π οΈ carefully extracting math data from Common Crawl;
π iteratively filtering and recalling high quality math pages using a classifier trained on synthetic annotations to identify math reasoning and deduction.
We conducted a series of ablations comparing the performance of Llama-3.2-3B-Base after continued pre-training on FineMath and observe notable gains compared to the baseline model and other public math datasets.
We hope this helps advance the performance of LLMs on math and reasoning! π
Weβre also releasing all the ablation models as well as the evaluation code.
HuggingFaceTB/finemath-6763fb8f71b6439b653482c2
Post
6829
We outperform Llama 70B with Llama 3B on hard math by scaling test-time compute π₯
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:
π Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time.
π 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.
π§ 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:
π Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time.
π 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.
π§ 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
2224
πͺπΊ Policy Thoughts in the EU AI Act Implementation πͺπΊ
There is a lot to like in the first draft of the EU GPAI Code of Practice, especially as regards transparency requirements. The Systemic Risks part, on the other hand, is concerning for both smaller developers and for external stakeholders.
I wrote more on this topic ahead of the next draft. TLDR: more attention to immediate large-scale risks and to collaborative solutions supported by evidence can help everyone - as long as developers disclose sufficient information about their design choices and deployment contexts.
Full blog here, based on our submitted response with @frimelle and @brunatrevelin :
https://huggingface.co/blog/yjernite/eu-draft-cop-risks#on-the-proposed-taxonomy-of-systemic-risks
There is a lot to like in the first draft of the EU GPAI Code of Practice, especially as regards transparency requirements. The Systemic Risks part, on the other hand, is concerning for both smaller developers and for external stakeholders.
I wrote more on this topic ahead of the next draft. TLDR: more attention to immediate large-scale risks and to collaborative solutions supported by evidence can help everyone - as long as developers disclose sufficient information about their design choices and deployment contexts.
Full blog here, based on our submitted response with @frimelle and @brunatrevelin :
https://huggingface.co/blog/yjernite/eu-draft-cop-risks#on-the-proposed-taxonomy-of-systemic-risks
- 2 replies
Post
5299
We are proud to announce
HuggingFaceFW/fineweb-2: A sparkling update to
HuggingFaceFW/fineweb with 1000s of π£οΈlanguages.
We applied the same data-driven approach that led to SOTA English performance inπ· FineWeb to thousands of languages.
π₯ FineWeb2 has 8TB of compressed text data and outperforms other multilingual datasets in our experiments.
The dataset is released under the permissive π ODC-By 1.0 license, and the π» code to reproduce it and our evaluations is public.
We will very soon announce a big community project, and are working on a π 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 inπ· FineWeb to thousands of languages.
π₯ FineWeb2 has 8TB of compressed text data and outperforms other multilingual datasets in our experiments.
The dataset is released under the permissive π ODC-By 1.0 license, and the π» code to reproduce it and our evaluations is public.
We will very soon announce a big community project, and are working on a π 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
1544
Exponentially growing number of open-source AI models over the course of the past 30 months β 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
1491
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
2262
Making SmolLM2 reproducible: open-sourcing our training & evaluation toolkit π οΈ https://github.com/huggingface/smollm/
- Pre-training code with nanotron
- Evaluation suite with lighteval
- Synthetic data generation using distilabel (powers our new SFT dataset HuggingFaceTB/smoltalk)
- Post-training scripts with TRL & the alignment handbook
- On-device tools with llama.cpp for summarization, rewriting & agents
Apache 2.0 licensed. V2 pre-training data mix coming soon!
Which other tools should we add next?
- Pre-training code with nanotron
- Evaluation suite with lighteval
- Synthetic data generation using distilabel (powers our new SFT dataset HuggingFaceTB/smoltalk)
- Post-training scripts with TRL & the alignment handbook
- On-device tools with llama.cpp for summarization, rewriting & agents
Apache 2.0 licensed. V2 pre-training data mix coming soon!
Which other tools should we add next?
Post
1716
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