Just ask users to vote with an emoji reaction for a specific option. Common in Discord. ;)
Bram Vanroy PRO
BramVanroy
AI & ML interests
Artificial intelligence, natural language processing, computational linguistics
Recent Activity
new activity 8 days ago
clips/beir-nl-hotpotqa:Licence field and card text differ (13 BEIR-NL subsets) liked a model 8 days ago
clips/e5-small-trm-nlOrganizations
reacted to Nymbo's post with 🔥 3 months ago
Post
2,620
Anthropic gave me six months of Claude Max 20x through the Claude for Open Source program, granted based on my Hugging Face work. Thank you 
Anthropic for supporting open source.
So far I've been pointing it at Markdown Minimap, an Obsidian plugin that adds a scrollable IDE-style minimap to your notes. This week I've been clearing a backlog of user-reported issues on it, with Claude often handling them end to end.
https://github.com/Nymbo/Markdown-Minimap — issues and PRs welcome.
So far I've been pointing it at Markdown Minimap, an Obsidian plugin that adds a scrollable IDE-style minimap to your notes. This week I've been clearing a backlog of user-reported issues on it, with Claude often handling them end to end.
https://github.com/Nymbo/Markdown-Minimap — issues and PRs welcome.
posted an update 3 months ago
Post
2,867
**I benchmarked HF buckets against https access for Common Crawl.**
Took me a while to get round to do this but I benchmarked access to Common Crawl via https vs hf buckets. Both experiments were run at night in Europe. I do not think other hardware problems were impacting the speeds since CPU processing time of the non-download pipeline components were highly similar (within 2% identical) and below only the WarcReader speeds of datatrove are used.
Experiment: selected 5 disjoint samples of 64 files each (randomly from the latest crawl; 20,499 docs/file). Those five batches were then processed by 32 single-core tasks with 4GB/core (five batches to calculate CIs). Paired experiment between using https and hf bucket.
- https: 40.0 [39.3-40.6] (seconds per WARC file)
- hf bucket: 172.0 [122.7-221.2]
That is a difference of about 4x in streaming speed. You'll see that https is also more stable (smaller CI).
I also ran raw throughput tests to the endpoints to measure rate limiting (64MiB transfer at 8/32/128/256 concurrent readers) and rate limiting seems not an issue for either: at any of those parallel reader numbers, their respective speeds stay about the same.
Note that, given CC scale, this is still a small test. Rate limiting may become more obvious when processing a full crawl. I do not know whether the https endpoint vs HF bucket will shut you out earlier with which limits.
Took me a while to get round to do this but I benchmarked access to Common Crawl via https vs hf buckets. Both experiments were run at night in Europe. I do not think other hardware problems were impacting the speeds since CPU processing time of the non-download pipeline components were highly similar (within 2% identical) and below only the WarcReader speeds of datatrove are used.
Experiment: selected 5 disjoint samples of 64 files each (randomly from the latest crawl; 20,499 docs/file). Those five batches were then processed by 32 single-core tasks with 4GB/core (five batches to calculate CIs). Paired experiment between using https and hf bucket.
- https: 40.0 [39.3-40.6] (seconds per WARC file)
- hf bucket: 172.0 [122.7-221.2]
That is a difference of about 4x in streaming speed. You'll see that https is also more stable (smaller CI).
I also ran raw throughput tests to the endpoints to measure rate limiting (64MiB transfer at 8/32/128/256 concurrent readers) and rate limiting seems not an issue for either: at any of those parallel reader numbers, their respective speeds stay about the same.
Note that, given CC scale, this is still a small test. Rate limiting may become more obvious when processing a full crawl. I do not know whether the https endpoint vs HF bucket will shut you out earlier with which limits.
reacted to prithivMLmods's post with 🚀 5 months ago
Post
5,660
I've made 8 Spaces in the Qwen-Image-Edit series, and out of them, 5 Spaces reached “Space of the Week”! A few Spaces are still topping the list even after many months.
Cumulatively, the series has crossed 8.2 million+ ZeroGPU runs and nearly 4 million visitors overall.
Thanks for all the community support! 🤗❤️
🔗 Spaces: https://huggingface.co/collections/prithivMLmods/image-generation-apps-collection
Cumulatively, the series has crossed 8.2 million+ ZeroGPU runs and nearly 4 million visitors overall.
Thanks for all the community support! 🤗❤️
🔗 Spaces: https://huggingface.co/collections/prithivMLmods/image-generation-apps-collection
- 4 replies
reacted to yuriyvnv's post with 🔥 6 months ago
Post
3,825
📄 The WAVe paper is officially out in the Information Sciences Journal.
You saw the PT and NL model releases earlier this year. This is the peer-reviewed paper behind them, with the full method, ablations, and downstream ASR evaluation.
Quick recap: WAVe is a 1B multimodal embedding model that filters synthetic speech at the word level, not the sentence level. On Portuguese ASR it cuts training steps by 34%, improves cross-domain generalization by 50%, and matches WER with 30% less synthetic data.
📦 Resources
- Paper: https://www.sciencedirect.com/science/article/pii/S0020025526005220
- PT model: yuriyvnv/WAVe-1B-Multimodal-PT
- NL model: yuriyvnv/WAVe-1B-Multimodal-NL
- Collection: https://huggingface.co/collections/yuriyvnv/multi-modal-embeddings-for-synthetic-transcript-filtering
- Code: https://github.com/yuriyvnv/WAVe
If you train ASR on synthetic or back-translated data, would like to see WAVe benchmarked on other languages.
@reach-vb @ylacombe @hf-audio @BramVanroy
#speech #asr #multimodal #syntheticdata #lowresource
You saw the PT and NL model releases earlier this year. This is the peer-reviewed paper behind them, with the full method, ablations, and downstream ASR evaluation.
Quick recap: WAVe is a 1B multimodal embedding model that filters synthetic speech at the word level, not the sentence level. On Portuguese ASR it cuts training steps by 34%, improves cross-domain generalization by 50%, and matches WER with 30% less synthetic data.
📦 Resources
- Paper: https://www.sciencedirect.com/science/article/pii/S0020025526005220
- PT model: yuriyvnv/WAVe-1B-Multimodal-PT
- NL model: yuriyvnv/WAVe-1B-Multimodal-NL
- Collection: https://huggingface.co/collections/yuriyvnv/multi-modal-embeddings-for-synthetic-transcript-filtering
- Code: https://github.com/yuriyvnv/WAVe
If you train ASR on synthetic or back-translated data, would like to see WAVe benchmarked on other languages.
@reach-vb @ylacombe @hf-audio @BramVanroy
#speech #asr #multimodal #syntheticdata #lowresource
reacted to yuriyvnv's post with 🚀 8 months ago
Post
620
🎯 WAVe-1B-Multimodal-NL: Word-Level Speech Quality Assessment for Dutch
Following the release of the Portuguese model, we're releasing the Dutch variant of WAVe — a 1B multimodal embedding model that assesses synthetic speech quality at the word level, thereby improving the quality of synthetically augmented datasets for training ASR models.
Trained on CommonVoice 16.1 Dutch with 5 corruption strategies, this model catches mispronunciations, timing errors, and prosody issues in synthetic data that sentence-level embeddings miss entirely.
Resources
- Dutch model: yuriyvnv/WAVe-1B-Multimodal-NL
- Portuguese model: yuriyvnv/WAVe-1B-Multimodal-PT
- Code: https://github.com/yuriyvnv/WAVe
This model builds on CommonVoice Dutch data — thanks to @mozilla and the CommonVoice community for making multilingual speech data accessible.
Would be great to hear from the Dutch NLP community — @BramVanroy @GroNLP — especially if you're working on Dutch ASR or TTS pipelines where quality filtering could help. Also tagging @hf-audio as this sits at the intersection of speech processing and data curation.
Following the release of the Portuguese model, we're releasing the Dutch variant of WAVe — a 1B multimodal embedding model that assesses synthetic speech quality at the word level, thereby improving the quality of synthetically augmented datasets for training ASR models.
Trained on CommonVoice 16.1 Dutch with 5 corruption strategies, this model catches mispronunciations, timing errors, and prosody issues in synthetic data that sentence-level embeddings miss entirely.
Resources
- Dutch model: yuriyvnv/WAVe-1B-Multimodal-NL
- Portuguese model: yuriyvnv/WAVe-1B-Multimodal-PT
- Code: https://github.com/yuriyvnv/WAVe
This model builds on CommonVoice Dutch data — thanks to @mozilla and the CommonVoice community for making multilingual speech data accessible.
Would be great to hear from the Dutch NLP community — @BramVanroy @GroNLP — especially if you're working on Dutch ASR or TTS pipelines where quality filtering could help. Also tagging @hf-audio as this sits at the intersection of speech processing and data curation.
reacted to onekq's post with 🔥 12 months ago
Post
2,886
The reaction on the QAT post is beyond expectations so below is my optimizer post as promised. But I found that I had lots of explanation to do about optimizer itself. So this post is actually a historical recount. The Muon optimizer (used by Kimi) post (coming very soon) can only continue after this.
https://huggingface.co/blog/onekq/adam-optimizer
If you know Adam(W) optimizer already, you can just skip and sorry for the wait. Otherwise, it should be a useful read.
https://huggingface.co/blog/onekq/adam-optimizer
If you know Adam(W) optimizer already, you can just skip and sorry for the wait. Otherwise, it should be a useful read.
posted an update about 1 year ago
Post
801
What are currently the best multilingual models with at most 72B parameters? Are Llama 3.3 70B and Qwen 2.5 72B still king?
- 1 reply
reacted to Reubencf's post with 🔥 about 1 year ago
Post
2,450
Releasing Version 1 of the Konkani LLM
Konkani is a low resource Indian language spoken by 2.5 Million people
Model:https://huggingface.co/Reubencf/gemma3-konkani
space: https://huggingface.co/spaces/Reubencf/Gemma3-konkani
Konkani is a low resource Indian language spoken by 2.5 Million people
Model:https://huggingface.co/Reubencf/gemma3-konkani
space: https://huggingface.co/spaces/Reubencf/Gemma3-konkani
- 2 replies
posted an update about 1 year ago
Post
1,183
Thanks to popular request, I've just added two subsets to the CommonCrawl-Creative Commons Corpus (C5; BramVanroy/CommonCrawl-CreativeCommons) so that you do not have to do filtering manually
- C5f ( BramVanroy/CommonCrawl-CreativeCommons-fine): only retains high-quality samples that are also present in FineWeb or FineWeb-2;
- C5r (https://huggingface.co/datasets/BramVanroy/CommonCrawl-CreativeCommons-recommended): additional strict filtering that removes samples with license disagreement, non-commercial licenses, and Wikipedia samples. The latter because you should probably get those from a more reliable source that provides better parsed content.
It goes without saying that these filters lead to a massive reduction in quantity. Doc and token counts are given on the dataset pages.
- C5f ( BramVanroy/CommonCrawl-CreativeCommons-fine): only retains high-quality samples that are also present in FineWeb or FineWeb-2;
- C5r (https://huggingface.co/datasets/BramVanroy/CommonCrawl-CreativeCommons-recommended): additional strict filtering that removes samples with license disagreement, non-commercial licenses, and Wikipedia samples. The latter because you should probably get those from a more reliable source that provides better parsed content.
It goes without saying that these filters lead to a massive reduction in quantity. Doc and token counts are given on the dataset pages.
reacted to pagezyhf's post with 🚀 over 1 year ago
Post
1,601
In our recent push to make more models available on Azure, we recently added SmolLM v3 in the catalog! 🚀
@juanjucm wrote a really detailed guide on how to deploy on Azure AI 🤗
https://huggingface.co/docs/microsoft-azure/azure-ai/examples/deploy-smollm3
If you want to see other models, please let us know
@juanjucm wrote a really detailed guide on how to deploy on Azure AI 🤗
https://huggingface.co/docs/microsoft-azure/azure-ai/examples/deploy-smollm3
If you want to see other models, please let us know
- 1 reply
replied to their post over 1 year ago
Special thanks to:
- The Common Crawl folks (Greg Lindahl, @pjox and others)
- The datatrove and FineWeb teams at Hugging Face (@guipenedo and others)
posted an update over 1 year ago
Post
3,795
📢💾 Introducing the Common Crawl Creative Commons Corpus (C5)!
C5 is a large-scale effort to heavily filter web-crawled data, as collected by the non-profit Common Crawl, to only documents that are Creative Commons-licensed such as cc-by-4.0 or public domain cc0. At this stage 150 billion tokens have been collected.
---
📄 data: BramVanroy/CommonCrawl-CreativeCommons
🧰 software: https://github.com/BramVanroy/CommonCrawl-CreativeCommons
---
</> To build C5, HTML pages are scrutinized and all links (if any) to CC licenses are collected, both in regular hyperlinks as well as in metadata. Additional data fields are included such as "was the license found in the
🌐 In this first version of C5, 8 languages are included (Afrikaans, German, English, French, Frysian, Italian, Dutch and Spanish). The language set was limited for two reasons: computational and storage limitations, and a collaboration with GPT-NL, which requested CC data for these languages to train a Dutch-focused, copyright-conscious LLM. In total, this V1 release contains almost 150 thousand documents and 150 billion tokens. This data was not filtered on quality nor deduplicated so that you can decide for yourself how much data to keep. To give some quality indication, a dataset field is present to describe whether a document is included in the FineWeb(-2) datasets, which are of high quality.
🔍 More work needs to be done! Only 7 out of 100+ Common Crawl crawls have been processed so far. That's encouraging because it means there is a lot more Creative Commons data to be collected! But to get there I need help in terms of compute. The current processing was already heavily sponsored by the Flemish Supercomputer but more is needed. If you have the compute available and which to collaborate in an open and transparent manner, please get in touch!
C5 is a large-scale effort to heavily filter web-crawled data, as collected by the non-profit Common Crawl, to only documents that are Creative Commons-licensed such as cc-by-4.0 or public domain cc0. At this stage 150 billion tokens have been collected.
---
📄 data: BramVanroy/CommonCrawl-CreativeCommons
🧰 software: https://github.com/BramVanroy/CommonCrawl-CreativeCommons
---
</> To build C5, HTML pages are scrutinized and all links (if any) to CC licenses are collected, both in regular hyperlinks as well as in metadata. Additional data fields are included such as "was the license found in the
head?" or "if multiple licenses were found, do they contradict each other?", which makes further filtering a breeze. 🌐 In this first version of C5, 8 languages are included (Afrikaans, German, English, French, Frysian, Italian, Dutch and Spanish). The language set was limited for two reasons: computational and storage limitations, and a collaboration with GPT-NL, which requested CC data for these languages to train a Dutch-focused, copyright-conscious LLM. In total, this V1 release contains almost 150 thousand documents and 150 billion tokens. This data was not filtered on quality nor deduplicated so that you can decide for yourself how much data to keep. To give some quality indication, a dataset field is present to describe whether a document is included in the FineWeb(-2) datasets, which are of high quality.
🔍 More work needs to be done! Only 7 out of 100+ Common Crawl crawls have been processed so far. That's encouraging because it means there is a lot more Creative Commons data to be collected! But to get there I need help in terms of compute. The current processing was already heavily sponsored by the Flemish Supercomputer but more is needed. If you have the compute available and which to collaborate in an open and transparent manner, please get in touch!
- 1 reply
reacted to julien-c's post with ❤️🔥 almost 2 years ago
Post
11,735
After some heated discussion 🔥, we clarify our intent re. storage limits on the Hub
TL;DR:
- public storage is free, and (unless blatant abuse) unlimited. We do ask that you consider upgrading to PRO and/or Enterprise Hub if possible
- private storage is paid above a significant free tier (1TB if you have a paid account, 100GB otherwise)
docs: https://huggingface.co/docs/hub/storage-limits
We optimize our infrastructure continuously to scale our storage for the coming years of growth in Machine learning, to the benefit of the community 🔥
cc: @reach-vb @pierric @victor and the HF team
TL;DR:
- public storage is free, and (unless blatant abuse) unlimited. We do ask that you consider upgrading to PRO and/or Enterprise Hub if possible
- private storage is paid above a significant free tier (1TB if you have a paid account, 100GB otherwise)
docs: https://huggingface.co/docs/hub/storage-limits
We optimize our infrastructure continuously to scale our storage for the coming years of growth in Machine learning, to the benefit of the community 🔥
cc: @reach-vb @pierric @victor and the HF team
- 29 replies
posted an update almost 2 years ago
Post
1,267
In the spirit of "Better late than never", I've finally written a brief overview paper for GEITje 7B Ultra. Initially released 10 months ago (oops), but still reaching around 1300 monthly downloads across the HF ecosystem (not including ollama).
GEITje 7B Ultra: A Conversational Model for Dutch (2412.04092)
While the paper discusses the model a little bit, I especially wanted to write about the datasets, which to this day seem an important asset for Dutch LLM training (SFT and preference tuning). We have a long way to go for Dutch, but publishing transparent and reproducible artefacts seems an important step to me, alongside having open discussions about data, bias, architectures.
In that spirit, thanks are in order for the creation of GEITje 7B Ultra and all related datasets:
- Michiel Buisman and UWV for providing the means to create the datasets
- Flemish Supercomputer Center (VSC) for the compute
- The Hugging Face Fellows and rest of the team for their discussions and insights
- The Dutch NLP community, notably @Rijgersberg for building the base GEITje model and the fruitful discussions we've had
More to come, step by step!
BramVanroy/geitje-7b-ultra-65c1ee010ad80fd1f6a8f208
GEITje 7B Ultra: A Conversational Model for Dutch (2412.04092)
While the paper discusses the model a little bit, I especially wanted to write about the datasets, which to this day seem an important asset for Dutch LLM training (SFT and preference tuning). We have a long way to go for Dutch, but publishing transparent and reproducible artefacts seems an important step to me, alongside having open discussions about data, bias, architectures.
In that spirit, thanks are in order for the creation of GEITje 7B Ultra and all related datasets:
- Michiel Buisman and UWV for providing the means to create the datasets
- Flemish Supercomputer Center (VSC) for the compute
- The Hugging Face Fellows and rest of the team for their discussions and insights
- The Dutch NLP community, notably @Rijgersberg for building the base GEITje model and the fruitful discussions we've had
More to come, step by step!
BramVanroy/geitje-7b-ultra-65c1ee010ad80fd1f6a8f208
reacted to davidberenstein1957's post with 🚀 about 2 years ago
Post
3,043
🚀 We will be generating a preference dataset for DPO/ORPO and cleaning it with AI feedback during our upcoming meetup!
In this session, we'll walk you through the essentials of building a distilabel pipeline by exploring two key use cases: cleaning an existing dataset and generating a preference dataset for DPO/ORPO. You’ll also learn how to make the most of AI feedback, integrating Argilla to gather human feedback and improve the overall data quality.
This session is perfect for you
- if you’re getting started with distilabel or synthetic data
- if you want to learn how to use LLM inference endpoints for **free**
- if you want to discover new functionalities
- if you want to provide us with new feedback
Sign up here: https://lu.ma/dt0c7jru
In this session, we'll walk you through the essentials of building a distilabel pipeline by exploring two key use cases: cleaning an existing dataset and generating a preference dataset for DPO/ORPO. You’ll also learn how to make the most of AI feedback, integrating Argilla to gather human feedback and improve the overall data quality.
This session is perfect for you
- if you’re getting started with distilabel or synthetic data
- if you want to learn how to use LLM inference endpoints for **free**
- if you want to discover new functionalities
- if you want to provide us with new feedback
Sign up here: https://lu.ma/dt0c7jru
posted an update over 2 years ago
Post
2,232
The InstructGPT paper mentions that they insert 10% pretraining data during SFT, which they find improves the effect of PPO (IIUC). Has anyone else done later ablations on this? I've only seen the inverse suggested, mixing in SFT data during pretraining.
- 2 replies
reacted to adamm-hf's post with ❤️ over 2 years ago
Post
2,228
cooking up something....anyone interested in a daily activity tracker for HF?
- 12 replies
posted an update over 2 years ago
Post
2,325
All my models seem to be plagued by infinite lists. When you ask a question that requires it to write a list, it most often keeps adding bullet points or enumeration. I am wondering whether this is a result of using chatty GPT-4 as DPO preferences. Any thoughts?
- 1 reply