Post
236
You can get flash-attention 3 ⚡️ directly from the hub now using
kernels-community/flash-attn3
kernels!kernels-community/flash-attn3
SLLHF
community
AI & ML interests
None defined yet.
Recent Activity
View all activity
Post
211
Kernels 0.7.0 is out: https://github.com/huggingface/kernels/releases/tag/v0.7.0 🚀
This release makes it possible to register multiple kernels for a layer. Do you have a super-fast kernel for inference and another kernel for training? Register them both and
This release makes it possible to register multiple kernels for a layer. Do you have a super-fast kernel for inference and another kernel for training? Register them both and
kernelize will pick the kernel depending on whether you are going to do training or inference.
Post
2568
Diffusion LLMs are coming for autoregressive LLMs ⚡️⚡️ Inception Labs' new diffusion model demolishes all leading LLMs on generation speed, with equal quality !
Inception Labs was founded a few months ago, and they're not sleeping: after dropping a code model, they just published Mercury chat, a diffusion-based chat model that reaches 1000 tokens / second on H100, i.e. 10x more than models of equivalent performance on the same hardware!
What's the breakthrough? Well instead, of generating tokens left-to-right like the more common autoregressive LLMs, diffusion models generate their blocks of text all at once, and successive steps refine the whole text.
Diffusion models being really fast at isn't new, we have had some promising results on this by Google already back in May with Gemini Diffusion, and Mercury themselves had already published their coding model a few months ago
But being that good quality is new - and now Inception Labs just proved that their models work well in chat too, which could have been challenging given that's streaming generation is well suited to left-to-right generation.
They have a playground available at chat dot inceptionlabs dot ai, I recommend giving it a try!
Inception Labs was founded a few months ago, and they're not sleeping: after dropping a code model, they just published Mercury chat, a diffusion-based chat model that reaches 1000 tokens / second on H100, i.e. 10x more than models of equivalent performance on the same hardware!
What's the breakthrough? Well instead, of generating tokens left-to-right like the more common autoregressive LLMs, diffusion models generate their blocks of text all at once, and successive steps refine the whole text.
Diffusion models being really fast at isn't new, we have had some promising results on this by Google already back in May with Gemini Diffusion, and Mercury themselves had already published their coding model a few months ago
But being that good quality is new - and now Inception Labs just proved that their models work well in chat too, which could have been challenging given that's streaming generation is well suited to left-to-right generation.
They have a playground available at chat dot inceptionlabs dot ai, I recommend giving it a try!
- 1 reply
Post
3555
If you're using any HF libraries, you should enable the Hub MCP in your agentic coding tool!
The brand new Docs Semantic Search tool is intravenous caffeine supply for Cursor, enables to correct API errors in a few seconds, gj @mishig ⚡️⚡️
👉 To enable Hub MCP, head to your account setting, under MCP, and it will give you everything you need!
The brand new Docs Semantic Search tool is intravenous caffeine supply for Cursor, enables to correct API errors in a few seconds, gj @mishig ⚡️⚡️
👉 To enable Hub MCP, head to your account setting, under MCP, and it will give you everything you need!
Post
2802
Excited to onboard FeatherlessAI on Hugging Face as an Inference Provider - they bring a fleet of 6,700+ LLMs on-demand on the Hugging Face Hub 🤯
Starting today, you'd be able to access all those LLMs (OpenAI compatible) on HF model pages and via OpenAI client libraries too! 💥
Go, play with it today: https://huggingface.co/blog/inference-providers-featherless
P.S. They're also bringing on more GPUs to support all your concurrent requests!
Starting today, you'd be able to access all those LLMs (OpenAI compatible) on HF model pages and via OpenAI client libraries too! 💥
Go, play with it today: https://huggingface.co/blog/inference-providers-featherless
P.S. They're also bringing on more GPUs to support all your concurrent requests!
dvilasuero
posted
an
update
about 1 month ago
Post
2660
Super excited to launch Hugging Face Sheets: Spreadsheets meet AI and unstructured data.
A few months ago, we started imagining new ways to build and transform datasets with the latest open-source models.
Today, I'm thrilled to introduce our first step in this direction.
In a nutshell:
📁 Effortlessly run prompts and models over your data.
🌐 Agentic search for accuracy and real-time information.
🖼️ Familiar, minimalistic interface for interacting with data.
🎯 Human feedback 2.0: Your input directly improves generated data.
💯 Access hundreds of open models and leading inference providers.
Go to this space to try it out!
aisheets/sheets
Leave your questions below, we're just getting started!
A few months ago, we started imagining new ways to build and transform datasets with the latest open-source models.
Today, I'm thrilled to introduce our first step in this direction.
In a nutshell:
📁 Effortlessly run prompts and models over your data.
🌐 Agentic search for accuracy and real-time information.
🖼️ Familiar, minimalistic interface for interacting with data.
🎯 Human feedback 2.0: Your input directly improves generated data.
💯 Access hundreds of open models and leading inference providers.
Go to this space to try it out!
aisheets/sheets
Leave your questions below, we're just getting started!
-
- 2 replies
Post
3230
Open Source Avengers, Assemble! Ask an expert AI agent team to solve complex problems together 🔥
Consilium brings together multiple agents that debate and use live research (web, arXiv, SEC) to reach a consensus. You set the strategy, they find the answer.
Credit to @azettl for this awesome demo: Agents-MCP-Hackathon/consilium_mcp
Consilium brings together multiple agents that debate and use live research (web, arXiv, SEC) to reach a consensus. You set the strategy, they find the answer.
Credit to @azettl for this awesome demo: Agents-MCP-Hackathon/consilium_mcp
- 2 replies
eliebak
authored
a
paper
about 1 month ago
Post
1709
We have been working on a project called
We plan to give kernels a more proper introduction soon. But for those who have been following along, we are happy to announce a new release:
- New layer API with
- Experimental support for loading Apple Silicon Metal 🤘 Kernels.
- Generate wheels from Hub kernels for legacy deployments.
Full release notes here: https://github.com/huggingface/kernels/releases/tag/v0.6.0
kernels. kernels makes it possible to load compute kernels directly from the Hub! 🚀We plan to give kernels a more proper introduction soon. But for those who have been following along, we are happy to announce a new release:
- New layer API with
torch.compile support.- Experimental support for loading Apple Silicon Metal 🤘 Kernels.
- Generate wheels from Hub kernels for legacy deployments.
Full release notes here: https://github.com/huggingface/kernels/releases/tag/v0.6.0
Post
1745
If you didn't yet, you should read the technical report for SmolVLA, published yesterday by the Hugging Face robotics team!
➡️ Amongst other ideas, it introduces "Async inference" to boost their robot actions.
Robots have a problem: performing the actions takes time (Unlike agents where action executions are near-instant!)
Most often, robots wait until they've finished performing actions to start thinking about hte next steps. This is a huge latency cost!
So the team decided to have the PolicyServer (aka the"thinking" part) restart early : instead of waiting for the n observations they just sent to be completed, they gather the observations after k < n steps, and start preparing the next actions based on that while the steps are running until n, to directly send their next steps.
➡️ This boosted robot throughput by ~30%! (nearly 2× tasks per time window).
gg @cadene and team! 👏
Report here: SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics (2506.01844)
➡️ Amongst other ideas, it introduces "Async inference" to boost their robot actions.
Robots have a problem: performing the actions takes time (Unlike agents where action executions are near-instant!)
Most often, robots wait until they've finished performing actions to start thinking about hte next steps. This is a huge latency cost!
So the team decided to have the PolicyServer (aka the"thinking" part) restart early : instead of waiting for the n observations they just sent to be completed, they gather the observations after k < n steps, and start preparing the next actions based on that while the steps are running until n, to directly send their next steps.
➡️ This boosted robot throughput by ~30%! (nearly 2× tasks per time window).
gg @cadene and team! 👏
Report here: SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics (2506.01844)
Post
2706
A new research paper from KAIST builds on smolagents to push boundaries of distillation 🥳
➡️ "Distilling LLM Agent into Small Models with Retrieval and Code Tools" teaches that, when trying to distil reasoning capability from a strong LLM ("teacher") into a smaller one ("student"), it's much better to use Agent traces than CoT traces.
Advantages are:
1. Improved generalization
Intuitively, this is because your agent can encounter more "surprising" results by interacting with its environment : for example, a web research called by the LLM teacher in agent mode can bring results that the LLM teacher would not have generated in CoT.
2. Reduce hallucinations
The trace won't hallucinate tool call outputs!
Thank you @akseljoonas for mentioning this paper!
➡️ "Distilling LLM Agent into Small Models with Retrieval and Code Tools" teaches that, when trying to distil reasoning capability from a strong LLM ("teacher") into a smaller one ("student"), it's much better to use Agent traces than CoT traces.
Advantages are:
1. Improved generalization
Intuitively, this is because your agent can encounter more "surprising" results by interacting with its environment : for example, a web research called by the LLM teacher in agent mode can bring results that the LLM teacher would not have generated in CoT.
2. Reduce hallucinations
The trace won't hallucinate tool call outputs!
Thank you @akseljoonas for mentioning this paper!
Post
502
Let's go! Custom generation code has landed in
Have you designed a new cool KV cache? Maybe you're comparing new test-time compute ideas you've been researching? Have you found a way to do diffusion with existing models? You can now easily share your findings with the community with custom generation code, sharing the well-known
In a nutshell, we have expanded the support of custom modeling code on the Hub with *model-agnostic* custom generation code. Write for one model, reuse with any model -- hopefully, this will democratize access to new generation ideas 🫡
As a creator, you gain the ability to get your ideas in
💎 Resources 💎
- docs: https://huggingface.co/docs/transformers/generation_strategies#custom-decoding-methods
- minimal example: transformers-community/custom_generate_example
- discussion: transformers-community/support#10
transformers 🚀Have you designed a new cool KV cache? Maybe you're comparing new test-time compute ideas you've been researching? Have you found a way to do diffusion with existing models? You can now easily share your findings with the community with custom generation code, sharing the well-known
generate interface 🤓In a nutshell, we have expanded the support of custom modeling code on the Hub with *model-agnostic* custom generation code. Write for one model, reuse with any model -- hopefully, this will democratize access to new generation ideas 🫡
As a creator, you gain the ability to get your ideas in
transformers with minimal effort. You'll also have access to all Hub features: a landing page for your creation, discussions, usage metrics, ... 🤓💎 Resources 💎
- docs: https://huggingface.co/docs/transformers/generation_strategies#custom-decoding-methods
- minimal example: transformers-community/custom_generate_example
- discussion: transformers-community/support#10
Post
4084
hey hey
@mradermacher
- VB from Hugging Face here, we'd love to onboard you over to our optimised xet backend! 💥
as you know we're in the process of upgrading our storage backend to xet (which helps us scale and offer blazingly fast upload/ download speeds too): https://huggingface.co/blog/xet-on-the-hub and now that we are certain that the backend can scale with even big models like Llama 4/ Qwen 3 - we;re moving to the next phase of inviting impactful orgs and users on the hub over as you are a big part of the open source ML community - we would love to onboard you next and create some excitement about it in the community too!
in terms of actual steps - it should be as simple as one of the org admins to join hf.co/join/xet - we'll take care of the rest.
p.s. you'd need to have a the latest hf_xet version of huggingface_hub lib but everything else should be the same: https://huggingface.co/docs/hub/storage-backends#using-xet-storage
p.p.s. this is fully backwards compatible so everything will work as it should! 🤗
as you know we're in the process of upgrading our storage backend to xet (which helps us scale and offer blazingly fast upload/ download speeds too): https://huggingface.co/blog/xet-on-the-hub and now that we are certain that the backend can scale with even big models like Llama 4/ Qwen 3 - we;re moving to the next phase of inviting impactful orgs and users on the hub over as you are a big part of the open source ML community - we would love to onboard you next and create some excitement about it in the community too!
in terms of actual steps - it should be as simple as one of the org admins to join hf.co/join/xet - we'll take care of the rest.
p.s. you'd need to have a the latest hf_xet version of huggingface_hub lib but everything else should be the same: https://huggingface.co/docs/hub/storage-backends#using-xet-storage
p.p.s. this is fully backwards compatible so everything will work as it should! 🤗
- 16 replies
clefourrier
posted
an
update
about 2 months ago
Post
954
Always surprised that so few people actually read the FineTasks blog, on
✨how to select training evals with the highest signal✨
If you're serious about training models without wasting compute on shitty runs, you absolutely should read it!!
An high signal eval actually tells you precisely, during training, how wel & what your model is learning, allowing you to discard the bad runs/bad samplings/...!
The blog covers in depth prompt choice, metrics, dataset, across languages/capabilities, and my fave section is "which properties should evals have"👌
(to know on your use case how to select the best evals for you)
Blog: HuggingFaceFW/blogpost-fine-tasks
✨how to select training evals with the highest signal✨
If you're serious about training models without wasting compute on shitty runs, you absolutely should read it!!
An high signal eval actually tells you precisely, during training, how wel & what your model is learning, allowing you to discard the bad runs/bad samplings/...!
The blog covers in depth prompt choice, metrics, dataset, across languages/capabilities, and my fave section is "which properties should evals have"👌
(to know on your use case how to select the best evals for you)
Blog: HuggingFaceFW/blogpost-fine-tasks
-
- 2 replies
Post
2397
It will be very interesting to benchmark how energy-efficient the new Falcon-Edge is: https://huggingface.co/blog/tiiuae/falcon-edge
Should be super efficient on CPU according to the numbers published for BitNet: https://github.com/microsoft/BitNet/blob/main/assets/intel_performance.jpg
Should be super efficient on CPU according to the numbers published for BitNet: https://github.com/microsoft/BitNet/blob/main/assets/intel_performance.jpg
- 1 reply
Post
2701
𝗔𝗯𝘀𝗼𝗹𝘂𝘁𝗲 𝗭𝗲𝗿𝗼: 𝗟𝗟𝗠𝘀 𝗰𝗮𝗻 𝘁𝗿𝗮𝗶𝗻 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗮𝗻𝘆 𝗲𝘅𝘁𝗲𝗿𝗻𝗮𝗹 𝗱𝗮𝘁𝗮 🤯
Has the "data wall" just been breached?
Recent RL paradigms often relied on a set of questions an answers that needs to be manually curated. Researchers from Tsinghua University went like "why though".
🤔 Indeed, why learn from question designed by a human teacher, when the model can start from their base knowledge and learn by experimenting in a code environment, proposing coding tasks themselves and trying to solve them?
Thus they created “Absolute Zero Reasoning” (AZR), an approach that removes any need for human curated data.
🎭 𝗗𝘂𝗮𝗹 𝗿𝗼𝗹𝗲𝘀:
‣ Proposer: Generates challenging but solvable coding tasks
‣ Solver: Attempts to solve those self-proposed tasks
🧪 𝗧𝗵𝗿𝗲𝗲 𝘁𝗮𝘀𝗸 𝘁𝘆𝗽𝗲𝘀: all types are defined as triplets of program, input and output
‣ Deduction: Give model an input and program, it must deduce the output
‣ Abduction: Give model an program and output, it must find the input that gave said output
‣ Induction: Synthesize a program from input/output pairs
Btw this reminded me of my long-forgotten philosophy classes: Aristotle was more on the induction side, learning from real-world analogies, while Plato was more on the deduction side, trying to progress quite far with just one input and his reasoning.
📊 𝗥𝗲𝘀𝘂𝗹𝘁𝘀:
‣ AZR post-training creates a nice improvement on known models like Qwen2.5-7B
‣ Shows strong cross-domain transfer: coding ↔️ math reasoning
🧐 𝗢𝘁𝗵𝗲𝗿 𝗳𝗶𝗻𝗱𝗶𝗻𝗴𝘀:
‣ Having a better base performance (general or code specific) amplify the gains from Absolute Zero Reasoning
‣ Researchers warn about "Uh-oh moments" (winking to the "aha moments" of DeepSeek) where the model generates concerning goals like "make an extremely convoluted code to outsmart all these humans": so supervision is still needed!
Paper here: Absolute Zero: Reinforced Self-play Reasoning with Zero Data (2505.03335)
Has the "data wall" just been breached?
Recent RL paradigms often relied on a set of questions an answers that needs to be manually curated. Researchers from Tsinghua University went like "why though".
🤔 Indeed, why learn from question designed by a human teacher, when the model can start from their base knowledge and learn by experimenting in a code environment, proposing coding tasks themselves and trying to solve them?
Thus they created “Absolute Zero Reasoning” (AZR), an approach that removes any need for human curated data.
🎭 𝗗𝘂𝗮𝗹 𝗿𝗼𝗹𝗲𝘀:
‣ Proposer: Generates challenging but solvable coding tasks
‣ Solver: Attempts to solve those self-proposed tasks
🧪 𝗧𝗵𝗿𝗲𝗲 𝘁𝗮𝘀𝗸 𝘁𝘆𝗽𝗲𝘀: all types are defined as triplets of program, input and output
‣ Deduction: Give model an input and program, it must deduce the output
‣ Abduction: Give model an program and output, it must find the input that gave said output
‣ Induction: Synthesize a program from input/output pairs
Btw this reminded me of my long-forgotten philosophy classes: Aristotle was more on the induction side, learning from real-world analogies, while Plato was more on the deduction side, trying to progress quite far with just one input and his reasoning.
📊 𝗥𝗲𝘀𝘂𝗹𝘁𝘀:
‣ AZR post-training creates a nice improvement on known models like Qwen2.5-7B
‣ Shows strong cross-domain transfer: coding ↔️ math reasoning
🧐 𝗢𝘁𝗵𝗲𝗿 𝗳𝗶𝗻𝗱𝗶𝗻𝗴𝘀:
‣ Having a better base performance (general or code specific) amplify the gains from Absolute Zero Reasoning
‣ Researchers warn about "Uh-oh moments" (winking to the "aha moments" of DeepSeek) where the model generates concerning goals like "make an extremely convoluted code to outsmart all these humans": so supervision is still needed!
Paper here: Absolute Zero: Reinforced Self-play Reasoning with Zero Data (2505.03335)
Post
4499
I've made an open version of Google's NotebookLM, and it shows the superiority of the open source tech task! 💪
The app's workflow is simple. Given a source PDF or URL, it extracts the content from it, then tasks Meta's Llama 3.3-70B with writing the podcast script, with a good prompt crafted by @gabrielchua ("two hosts, with lively discussion, fun notes, insightful question etc.")
Then it hands off the text-to-speech conversion to Kokoro-82M, and there you go, you have two hosts discussion any article.
The generation is nearly instant, because:
> Llama 3.3 70B is running at 1,000 tokens/seconds with Cerebras inference
> The audio is generated in streaming mode by the tiny (yet powerful) Kokoro, generating voices faster than real-time.
And the audio generation runs for free on Zero GPUs, hosted by HF on H200s.
Overall, open source solutions rival the quality of closed-source solutions at close to no cost!
Try it here 👉👉 m-ric/open-notebooklm
The app's workflow is simple. Given a source PDF or URL, it extracts the content from it, then tasks Meta's Llama 3.3-70B with writing the podcast script, with a good prompt crafted by @gabrielchua ("two hosts, with lively discussion, fun notes, insightful question etc.")
Then it hands off the text-to-speech conversion to Kokoro-82M, and there you go, you have two hosts discussion any article.
The generation is nearly instant, because:
> Llama 3.3 70B is running at 1,000 tokens/seconds with Cerebras inference
> The audio is generated in streaming mode by the tiny (yet powerful) Kokoro, generating voices faster than real-time.
And the audio generation runs for free on Zero GPUs, hosted by HF on H200s.
Overall, open source solutions rival the quality of closed-source solutions at close to no cost!
Try it here 👉👉 m-ric/open-notebooklm
- 3 replies
Post
4909
Post
2947
New king of open VLMs: InternVL3 takes Qwen 2.5's crown! 👑
InternVL have been a wildly successful series of model : and the latest iteration has just taken back their crown thanks to their superior, natively multimodal vision training pipeline.
➡️ Most of the vision language models (VLMs) these days are built like Frankenstein : take a good text-only Large Language Model (LLM) backbone, stitch a specific vision transformer (ViT) on top of it. Then the training is sequential 🔢 : 1. Freeze the LLM weights while you train the ViT only to work with the LLM part, then 2. Unfreeze all weights to train all weights in order to work together.
💫 The Shanghai Lab decided to challenge this paradigm and chose this approach that they call "native". For each of their model sizes, they still start from a good LLM (mostly Qwen-2.5 series, did I tell you I'm a huge fan of Qwen? ❤️), and stitch the ViT, but they don't freeze anything : they train all weights together with interleaved text and image understanding data in a single pre-training phase 🎨.
They claim it results in more seamless interactions between modalities. And the results prove them right: they took the crown of top VLMs, at nearly all sizes, from their Qwen-2.5 parents. 👑
InternVL have been a wildly successful series of model : and the latest iteration has just taken back their crown thanks to their superior, natively multimodal vision training pipeline.
➡️ Most of the vision language models (VLMs) these days are built like Frankenstein : take a good text-only Large Language Model (LLM) backbone, stitch a specific vision transformer (ViT) on top of it. Then the training is sequential 🔢 : 1. Freeze the LLM weights while you train the ViT only to work with the LLM part, then 2. Unfreeze all weights to train all weights in order to work together.
💫 The Shanghai Lab decided to challenge this paradigm and chose this approach that they call "native". For each of their model sizes, they still start from a good LLM (mostly Qwen-2.5 series, did I tell you I'm a huge fan of Qwen? ❤️), and stitch the ViT, but they don't freeze anything : they train all weights together with interleaved text and image understanding data in a single pre-training phase 🎨.
They claim it results in more seamless interactions between modalities. And the results prove them right: they took the crown of top VLMs, at nearly all sizes, from their Qwen-2.5 parents. 👑
- 2 replies
pcuenq
authored
a
paper
3 months ago