badayvedat (Vedat Baday)

Hi Huggingfacers!

Thrilled to introduce Adam-mini, an optimizer that achieves on-par or better performance than AdamW with 45% to 50% less memory footprint. Adam-mini can also achieve 49.5% higher throughput than AdamW on Llama2-7B pre-training.

The design of Adam-mini is inspired by certain Hessian structures we observed on Transformers.

Feel free to try it out! Try switching to Adam-mini with the same hyperparams of AdamW, it would work with only half memory. Hope Adam-mini can help save time, cost, and energy in your tasks!

Paper: "Adam-mini: Use Fewer Learning Rates To Gain More" https://arxiv.org/abs/2406.16793

Code: https://github.com/zyushun/Adam-mini

1 reply

·

Reacted to Xenova's post with 🔥 5 months ago

Post

5959

Florence-2, the new vision foundation model by Microsoft, can now run 100% locally in your browser on WebGPU, thanks to Transformers.js! 🤗🤯

It supports tasks like image captioning, optical character recognition, object detection, and many more! 😍 WOW!
- Demo: Xenova/florence2-webgpu
- Models: https://huggingface.co/models?library=transformers.js&other=florence2
- Source code: https://github.com/xenova/transformers.js/tree/v3/examples/florence2-webgpu

liked a model 5 months ago

fal/AuraSR

Updated Jul 15 • 755 • 294

liked a Space 7 months ago

Running

81

📊

imgsys.org

imgsys.org -- arena for text guided image generation

Reacted to isidentical's post with ❤️ 7 months ago

Post

2096

Happy to announce https://imgsys.org -- a sister project to Chatbot Arena by lmsys -- for comparing different text guided image generation models models. Try it natively on HuggingFace: https://huggingface.co/spaces/fal-ai/imgsys

1 reply

·

Reacted to wanghaofan's post with 🔥 8 months ago

Post

3257

Greeting!

We are happy to introduce our InstantStyle, which is a framework that employs straightforward yet potent techniques for achieving effective disentanglement of style and content from reference images.

Code: https://github.com/InstantStyle/InstantStyle
Paper: InstantStyle: Free Lunch towards Style-Preserving in Text-to-Image Generation (2404.02733)
Project Page: https://instantstyle.github.io/

Reacted to Jaward's post with ❤️ 8 months ago

Post

2827

After giving GPU Programming a hands-on try, I have come to appreciate the level of complexity in AI compute:

- Existing/leading frameworks (CUDA, OpenCL, DSLs, even Triton), still fall at the mercy of low-level compute that requires deeper understanding and experience.
- Ambiguous optimizations methods that will literally drive you mad 🤯
- Triton is cool but not cool enough (high level abstractions that fall back to low level compute issues as you build more specialized kernels)
- As for CUDA, optimization requires considering all major components of the GPU (DRAM, SRAM, ALUs) 🤕
- Models today require stallion written GPU kernels to reduce storage and compute cost.
- GPTQ was a big save 👍🏼

@karpathy is right expertise in this area is scarce and the reason is quite obvious - uncertainties: we are still struggling to get peak performance from multi-connected GPUs while maintaining precision and reducing cost.

May the Scaling Laws favor us lol.

5 replies

·

New activity in ByteDance/AnimateDiff-Lightning 8 months ago

Reproducing ComfyUI examples

1

#6 opened 8 months ago by

badayvedat

New activity in multimodalart/stable-cascade 8 months ago

Something's broken

5

#16 opened 8 months ago by

FireCat1

Reacted to malhajar's post with ❤️ 9 months ago

Post

🚀 Major Update: OpenLLM Turkish Benchmarks & Leaderboard Launch! 🚀

Exciting news for the Hugging Face community! I'm thrilled to announce the launch of my fully translated OpenLLM Benchmarks in Turkish, accompanied by my innovative leaderboard, ready to highlight the capabilities of Turkish language models. This marks a landmark achievement in supporting and advancing Turkish AI research.

What’s New:

📚 Complete OpenLLM Benchmarks in Turkish: Dive into my comprehensive suite of benchmarks, now available for thorough evaluation of Turkish LLMs.

📈 Live Leaderboard: Explore my live leaderboard showcasing the progress and excellence in Turkish language AI. (Note: Current evaluations are conducted manually but are consistently updated.)

Partnership Invitation:

🤝 Join My Automation Mission: I'm on the lookout for partners to help transition from manual to automated leaderboard evaluations. Your support can catalyze real-time, streamlined assessments, pushing Turkish LLMs to new heights.
Key Resources:

📚 Explore the Turkish OpenLLM Collection: ( malhajar/openllmturkishleadboard-datasets-65e5854490a87c0f2670ec18)

🏆 Discover the Leaderboard: ( malhajar/OpenLLMTurkishLeaderboard)

Get Involved:

💡 Share Your Models: Contribute to the burgeoning field of Turkish AI, showcasing your work and contributing to the collective progress.

Let's unite to propel Turkish AI forward and set a precedent for the global community. Stay tuned as I plan to expand these efforts to other languages, further enriching the AI ecosystem!

Join this groundbreaking endeavor and let’s shape the future of AI together! 🌐

#TurkishLLM #AI #MachineLearning #LanguageModels #OpenLLM #HuggingFace