fusing ((De)fusing)

sayakpaul

posted an update 16 days ago

Post

898

Fast LoRA inference for Flux with Diffusers and PEFT 🚨

There are great materials that demonstrate how to optimize inference for popular image generation models, such as Flux. However, very few cover how to serve LoRAs fast, despite LoRAs being an inseparable part of their adoption.

In our latest post, @BenjaminB and I show different techniques to optimize LoRA inference for the Flux family of models for image generation. Our recipe includes the use of:

1. torch.compile
2. Flash Attention 3 (when compatible)
3. Dynamic FP8 weight quantization (when compatible)
4. Hotswapping for avoiding recompilation during swapping new LoRAs 🤯

We have tested our recipe with Flux.1-Dev on both H100 and RTX 4090. We achieve at least a *2x speedup* in either of the GPUs. We believe our recipe is grounded in the reality of how LoRA-based use cases are generally served. So, we hope this will be beneficial to the community 🤗

Even though our recipe was tested primarily with NVIDIA GPUs, it should also work with AMD GPUs.

Learn the details and the full code here:
https://huggingface.co/blog/lora-fast

thomwolf

authored a paper about 1 month ago

FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language

Paper • 2506.20920 • Published Jun 26 • 64

multimodalart

posted an update about 2 months ago

Post

9309

Self-Forcing - a real-time video distilled model from Wan 2.1 by @adobe is out, and they open sourced it 🐐

I've built a live real time demo on Spaces 📹💨

multimodalart/self-forcing

5 replies

·

thomwolf

authored a paper 2 months ago

SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics

Paper • 2506.01844 • Published Jun 2 • 126

sayakpaul

posted an update 3 months ago

Post

2775

Diffusers supports a good variety of quantization backends. It can be challenging to navigate through them, given the complex nature of diffusion pipelines in general.

So, @derekl35 set out to write a comprehensive guide that puts users in the front seat. Explore the different backends we support, learn the trade-offs they offer, and finally, check out the cool space we built that lets you compare quantization results.

Give it a go here:
https://lnkd.in/gf8Pi4-2

sayakpaul

posted an update 3 months ago

Post

1746

Despite the emergence of combining LLM and DiT architectures for T2I synthesis, its design remains severely understudied.

This was done long ago and got into CVPR25 -- super excited to finally share it now, along with the data and code ♥️

We explore several architectural choices that affect this design. We provide an open & reproducible training recipe that works at scale.

Works like Playground v3 have already explored a deep fusion between an LLM and a DiT, sharing their representations through layerwise attention. They exhibit excellent performance on T2I.

Despite its compelling results and other performance virtues, it remains unexplored, which is what we want to improve in our work. Specifically, we take a pre-trained LLM (Gemma-2B) and trainable DiT, and set out to explore what makes a "good deep fusion" between the two for T2I.

We explore several key questions in the work, such as:

Q1: How should we do attention? We considered several alternatives. PixArt-Alpha like attention (cross-attention) is very promising.
Q2: Should we incorporate additional text modulation?
Q3: Can we eliminate timestep conditioning?
Q4: How do we do positional encodings?
Q5: Do instruction-tuned LLMs help deep fusion?
Q6: Would using a decoder LLM from a multimodal model be helpful?
Q7: Does using a better variant of Gemma help?

Based on the above findings, we arrive at FuseDiT with the following components on top of the base architecture from the findings of our experiments.

* No AdaLN-Zero modules
* 1D + 2D-RoPE
* Gemma 2 2B, adjusting DiT configurations accordingly

We trained FuseDiT on a mixture from CC12M, JourneyDB, & SA (~26M image-text pairs) for 800 steps. While not the best model, it's encouraging to develop something in a guided manner using open datasets.

To know more (code, models, all are available), please check out the paper:
https://lnkd.in/gg6qyqZX.

sayakpaul

authored a paper 3 months ago

Exploring the Deep Fusion of Large Language Models and Diffusion Transformers for Text-to-Image Synthesis

Paper • 2505.10046 • Published May 15 • 9

sayakpaul

authored a paper 4 months ago

From Reflection to Perfection: Scaling Inference-Time Optimization for Text-to-Image Diffusion Models via Reflection Tuning

Paper • 2504.16080 • Published Apr 22 • 15

thomwolf

posted an update 4 months ago

Post

6229

If you've followed the progress of robotics in the past 18 months, you've likely noticed how robotics is increasingly becoming the next frontier that AI will unlock.

At Hugging Face—in robotics and across all AI fields—we believe in a future where AI and robots are open-source, transparent, and affordable; community-built and safe; hackable and fun. We've had so much mutual understanding and passion working with the Pollen Robotics team over the past year that we decided to join forces!

You can already find our open-source humanoid robot platform Reachy 2 on the Pollen website and the Pollen community and people here on the hub at

pollen-robotics

We're so excited to build and share more open-source robots with the world in the coming months!

1 reply

·

thomwolf

authored a paper 4 months ago

SmolVLM: Redefining small and efficient multimodal models

Paper • 2504.05299 • Published Apr 7 • 197

anton-l

authored a paper 4 months ago

SmolVLM: Redefining small and efficient multimodal models

Paper • 2504.05299 • Published Apr 7 • 197

pcuenq

authored a paper 4 months ago

SmolVLM: Redefining small and efficient multimodal models

Paper • 2504.05299 • Published Apr 7 • 197

thomwolf

authored a paper 4 months ago

YourBench: Easy Custom Evaluation Sets for Everyone

Paper • 2504.01833 • Published Apr 2 • 22

thomwolf

posted an update 4 months ago

Post

3714

The new DeepSite space is really insane for vibe-coders
enzostvs/deepsite

With the wave of vibe-coding-optimized LLMs like the latest open-source DeepSeek model (version V3-0324), you can basically prompt out-of-the-box and create any app and game in one-shot.

It feels so powerful to me, no more complex framework or under-the-hood prompt engineering to have a working text-to-app tool.

AI is eating the world and *open-source* AI is eating AI itself!

PS: and even more meta is that the DeepSite app and DeepSeek model are both fully open-source code => time to start recursively improve?

PPS: you still need some inference hosting unless you're running the 600B param model at home, so check the very nice list of HF Inference Providers for this model: deepseek-ai/DeepSeek-V3-0324

1 reply

·

sayakpaul

authored a paper 5 months ago

SANA-Sprint: One-Step Diffusion with Continuous-Time Consistency Distillation

Paper • 2503.09641 • Published Mar 12 • 40

thomwolf

posted an update 5 months ago

Post

3061

We've kept pushing our Open-R1 project, an open initiative to replicate and extend the techniques behind DeepSeek-R1.

And even we were mind-blown by the results we got with this latest model we're releasing: ⚡️OlympicCoder ( open-r1/OlympicCoder-7B and open-r1/OlympicCoder-32B)

It's beating Claude 3.7 on (competitive) programming –a domain Anthropic has been historically really strong at– and it's getting close to o1-mini/R1 on olympiad level coding with just 7B parameters!

And the best part is that we're open-sourcing all about its training dataset, the new IOI benchmark, and more in our Open-R1 progress report #3: https://huggingface.co/blog/open-r1/update-3

Datasets are are releasing:
- open-r1/codeforces
- open-r1/codeforces-cots
- open-r1/ioi
- open-r1/ioi-test-cases
- open-r1/ioi-sample-solutions
- open-r1/ioi-cots
- open-r1/ioi-2024-model-solutions

sayakpaul

posted an update 6 months ago

Post

3904

Inference-time scaling meets Flux.1-Dev (and others) 🔥

Presenting a simple re-implementation of "Inference-time scaling diffusion models beyond denoising steps" by Ma et al.

I did the simplest random search strategy, but results can potentially be improved with better-guided search methods.

Supports Gemini 2 Flash & Qwen2.5 as verifiers for "LLMGrading" 🤗

The steps are simple:

For each round:

1> Starting by sampling 2 starting noises with different seeds.
2> Score the generations w.r.t a metric.
3> Obtain the best generation from the current round.

If you have more compute budget, go to the next search round. Scale the noise pool (2 ** search_round) and repeat 1 - 3.

This constitutes the random search method as done in the paper by Google DeepMind.

Code, more results, and a bunch of other stuff are in the repository. Check it out here: https://github.com/sayakpaul/tt-scale-flux/ 🤗

anton-l

authored a paper 6 months ago

SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model

Paper • 2502.02737 • Published Feb 4 • 241

thomwolf

authored a paper 6 months ago

SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model

Paper • 2502.02737 • Published Feb 4 • 241

sayakpaul

posted an update 6 months ago

Post

2124

We have been cooking a couple of fine-tuning runs on CogVideoX with finetrainers, smol datasets, and LoRA to generate cool video effects like crushing, dissolving, etc.

We are also releasing a LoRA extraction utility from a fully fine-tuned checkpoint. I know that kind of stuff has existed since eternity, but the quality on video models was nothing short of spectacular. Below are some links:

* Models and datasets: