John6666 (John Smith)

reacted to mervenoyan's post with 🤗 about 5 hours ago

Post

217

Why do people sleep on DSE multimodal retrieval models? 👀

They're just like ColPali, but highly scalable, fast and you can even make them more efficient with binarization or matryoshka with little degradation 🪆

I made a small collection of them so you can get started merve/multimodal-dse-retrievers-67fe71a9c8f1ad26a48859c3

Image taken from MCDSE blog https://huggingface.co/blog/marco/announcing-mcdse-2b-v1

reacted to tomaarsen's post with 🔥 about 5 hours ago

Post

350

I just released Sentence Transformers v4.1; featuring ONNX and OpenVINO backends for rerankers offering 2-3x speedups and improved hard negatives mining which helps prepare stronger training datasets. Details:

🏎️ ONNX, OpenVINO, Optimization, Quantization
- I've added ONNX and OpenVINO support with just one extra argument: "backend" when loading the CrossEncoder reranker, e.g.: CrossEncoder("cross-encoder/ms-marco-MiniLM-L6-v2", backend="onnx")
- The export_optimized_onnx_model, export_dynamic_quantized_onnx_model, and export_static_quantized_openvino_model functions now work with CrossEncoder rerankers, allowing you to optimize (e.g. fusions, gelu approximations, etc.) or quantize (int8 weights) rerankers.
- I've uploaded ~340 ONNX & OpenVINO models for all existing models under the cross-encoder Hugging Face organization. You can use these without having to export when loading.

⛏ Improved Hard Negatives Mining
- Added 'absolute_margin' and 'relative_margin' arguments to mine_hard_negatives.
- absolute_margin ensures that sim(query, negative) < sim(query, positive) - absolute_margin, i.e. an absolute margin between the negative & positive similarities.
- relative_margin ensures that sim(query, negative) < sim(query, positive) * (1 - relative_margin), i.e. a relative margin between the negative & positive similarities.
- Inspired by the excellent NV-Retriever paper from NVIDIA.

And several other small improvements. Check out the full release notes here: https://github.com/UKPLab/sentence-transformers/releases/tag/v4.1.0

With this release, I introduce near-feature parity between the SentenceTransformer embedding & CrossEncoder reranker models, which I've wanted to do for quite some time! With rerankers very strongly supported now, it's time to look forward to other useful architectures!

reacted to S-Dreamer's post with 🔥 about 5 hours ago

Post

335

🚀 Introducing CyberForge by S-Dreamer!
CyberForge is our brand new Hugging Face Space that unites cybersecurity expertise with AI-driven innovation. Explore interactive demos, contribute code, and join a thriving community of researchers and developers dedicated to building smarter, scalable cybersecurity solutions.
👉 S-Dreamer/CyberForge
Let's collaborate, innovate, and fortify our digital future together!

reacted to bartowski's post with 👍 about 5 hours ago

Post

686

Access requests enabled for latest GLM models

While a fix is being implemented (https://github.com/ggml-org/llama.cpp/pull/12957) I want to leave the models up for visibility and continued discussion, but want to prevent accidental downloads of known broken models (even though there are settings that could fix it at runtime for now)

With this goal, I've enabled access requests. I don't really want your data, so I'm sorry that I don't think there's a way around that? But that's what I'm gonna do for now, and I'll remove the gate when a fix is up and verified and I have a chance to re-convert and quantize!

Hope you don't mind in the mean time :D

reacted to onekq's post with 👍 about 14 hours ago

Post

1018

I used three posts to explain GPU/CPU and LLM performances, now finally circle back to my own model.😅

OneSQL needs GPU because it processes long prompt. It is not a chatbot which replies short prompts with long answers. I call models of my kind workhorse models.

We all have to scramble for GPUs to get adoption. Below are a few ways.

You can inherit it. If you have a new Mac machine. Congratulations, you have GPU.

You can leverage it. Get inference providers to adopt your model, then you switch from CapEx to OpEx.

Or you buy it. Go frugal. Find older GPUs with enough HBMs to house your model.

reacted to merterbak's post with 🚀 about 14 hours ago

Post

1722

OpenAI published 2 benchmark datasets on Hugging Face 🔥
openai/mrcr
openai/graphwalks
MRCR tests how well a model can find the right answer when many similar questions are spread out in a long context. Graphwalks checks if a model can follow steps in a big graph and find the correct nodes by thinking through the structure

reacted to AdinaY's post with 👍 about 14 hours ago

Post

695

🔥 Big day for the Chinese open source AI community:

zh-ai-community

> Skywork AI :
Released 7B/32B reasoning models excels in math & coding
Skywork/skywork-or1-67fa1bcb41b436ef2def76b9

> Moonshot AI & Numina:
Dropped 1.5B/7B POWERFUL formal math reasoning models
AI-MO/kimina-prover-preview-67fb536b883d60e7ca25d7f9

> Zhipu AI :
Launched 9B/32B reasoning models powering their first general AI agent - AutoGLM ✨
THUDM/glm-4-0414-67f3cbcb34dd9d252707cb2e

> DeepSeek :
Announced to open source its internal inference engine: DeepSeek Inference Engine
https://github.com/deepseek-ai/open-infra-index/blob/main/OpenSourcing_DeepSeek_Inference_Engine/README.md

Can't wait for more exciting releases coming 🥳

1 reply

·

reacted to stefan-french's post with 😎 about 14 hours ago

Post

1525

🚀 We just released the latest version of our any-agent library, now with support for the Agno agent framework.
Check it out: https://github.com/mozilla-ai/any-agent

📖 Read more about the motivation behind the any-agent library in our blog post here: https://huggingface.co/blog/mozilla-ai/introducing-any-agent

reacted to DualityAI-RebekahBogdanoff's post with 👍 about 14 hours ago

Post

2238

We’re back—with higher stakes, new datasets, and more chances to stand out. Duality AI's Synthetic-to-Real Object Detection Challenge 2 is LIVE!🚦

✍ Sign up here: https://lnkd.in/g2avFP_X

After the overwhelming response to Challenge 1, we're pushing the boundaries even further in Challenge 2, where your object detection models will be put to the test in the real world after training only on synthetic data.

👉 Join our Synthetic-to-Real Object Detection Challenge 2 on Kaggle!

What’s Different This Time? Unlike our first challenge, we’re now diving deep into data manipulation. Competitors can:

🔹Access 4 new supplemental datasets via FalconCloud with varying lighting, occlusions, and camera angles.
🔹Generate your own synthetic datasets using FalconEditor to simulate edge cases.
🔹Mix, match, and build custom training pipelines for maximum mAP@50 performance

This challenge isn’t just about using synthetic data—it’s about mastering how to craft the right synthetic data.
Ready to test your skills?

🏆The Challenge
Train an object detection model using synthetic images created with Falcon—Duality AI's cutting-edge digital twin simulation software—then evaluate your model on real-world imagery.

The Twist?

📈Boost your model’s accuracy by creating and refining your own custom synthetic datasets using Falcon!

Win Cash Prizes & Recognition
🔹Earn cash and public shout-outs from the Duality AI accounts
Enhance Your Portfolio
🔹Demonstrate your real-world AI and ML expertise in object detection to prospective employers and collaborators.
🔹Expand Your Network
🔹Engage, compete, and collaborate with fellow ML engineers, researchers, and students.
🚀 Put your skills to the test and join our Kaggle competition today: https://lnkd.in/g2avFP_X

reacted to ImranzamanML's post with 🔥 about 14 hours ago

Post

2109

Mental Health Chatbot by Fine-Tuning Llama 4

https://huggingface.co/blog/ImranzamanML/llama-4-fine-tuning-with-mental-health-counseling

4 replies

·

reacted to JLouisBiz's post with 👀 1 day ago

Post

431

If you are using llama.CPP then From time to time you may have a need to quickly review your HTML output. And there is no automatic way to do it in its native web UI. This small shell script can help you integrate with your copy function. Just press on copy and invoke the shell script. You can make a small icon to invoke the shell script or bind it to the key or mouse button.

Shell script is here:

https://gitea.com/gnusupport/LLM-Helpers/src/branch/main/bin/clipboard-to-firefox.sh

And video demonstration is here: https://www.youtube.com/watch?v=WCu3TazXpgg

Join my Discord for LLM integration: https://discord.gg/N2BRPZ2jKb

reacted to AdinaY's post with ❤️ 1 day ago

Post

2207

🔥 New reasoning models from the Chinese community, by Skywork 天工-昆仑万维

Skywork/skywork-or1-67fa1bcb41b436ef2def76b9

✨Skywork OR1-Math-7B > Optimized for math reasoning
✨Skywork-OR1-7B-preview > Excels in math & coding
✨Skywork-OR1-32B-preview > Matches Deepseek-R1 on math (AIME24/25) and coding (LiveCodeBench)

Released under the Apache 2.0 license 🥳
Final version coming in 2 weeks!

reacted to merve's post with 🔥 1 day ago

Post

2694

sooo many open AI releases past week, let's summarize! 🤗
merve/april-11-releases-67fcd78be33d241c0977b9d2

multimodal
> Moonshot AI released Kimi VL Thinking, first working open-source multimodal reasoning model and Kimi VL Instruct, both 16B MoEs with 3B active params (OS)
> InternVL3 released based on Qwen2.5VL, 7 ckpts with various sizes (1B to 78B)

LLMs
> NVIDIA released Llama-3_1-Nemotron-Ultra-253B-v1 an LLM built on Llama 405B for reasoning, chat and tool use
> Agentica released DeepCoder-14B-Preview, fine-tuned version of DeepSeek-R1-Distilled-Qwen-14B on problem-test pairs, along with the compiled dataset
> Zyphra/ZR1-1.5B is a new small reasoning LLM built on R1-Distill-1.5B (OS)
> Skywork-OR1-32B-Preview is a new reasoning model by Skywork

Image Generation
> HiDream releases three new models, HiDream I1 Dev, I1 Full, and I1 fast for image generation (OS)

*OS ones have Apache 2.0 or MIT licenses

3 replies

·

reacted to Jaward's post with 👀 1 day ago

Post

1009

Funtime with SpatialLM- eventually it will serve well in embodied AI.

reacted to thomwolf's post with 🚀 1 day ago

Post

2404

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

·

reacted to leonardlin's post with 👍 1 day ago

Post

1734

Happy to announce the release of Shisa V2, our latest generation of our bilingual Japanese-English language models. After hundreds of ablations and months of work, we're releasing some of the strongest open Japanese models at 7B, 8B, 12B, 14B, 32B and 70B! Full announcement here https://shisa.ai/posts/shisa-v2/ or visit the Shisa V2 HF collection: shisa-ai/shisa-v2-67fc98ecaf940ad6c49f5689

reacted to MrDragonFox's post with 👍 1 day ago

Post

2523

yet a other audio datasets pre classified for events + audio aestetics

this time for german - 680h sampled from emilia yodas

timestamps for asr training or other fancier things available as nc in the raw repo

MrDragonFox/DE_Emilia_Yodas_680h

cc by 4.0 as by emilia yodas

raw events / transcriptions are cc by NC 4.0

MrDragonFox/DE_Emilia_Yodas_680h_raw_timestamps

the coming days i should push about 600h english + some japanese too same format

reacted to openfree's post with 🔥 1 day ago

Post

4441

Agentic AI Era: Analyzing MCP vs MCO 🚀

Hello everyone!
With the rapid advancement of AI agent technology, two architectures have come into the spotlight: MCP (Model Context Protocol) and MCO (Model Context Open-json). Today, we’ll introduce the key features and differences of these two approaches.

VIDraft/Agentic-AI-CHAT

MCP: The Traditional Approach 🏛️
Centralized Function Registry: All functions are hardcoded into the core system.

Static Function Definitions & Tight Coupling: New features require changes to the core application code, limiting scalability.

Monolithic Design: Complex deployment and version management can cause a single error to affect the whole system.

Code Example:
'''py
FUNCTION_REGISTRY = {
"existing_function": existing_function,
"new_function": new_function # Adding a new function
}
'''

MCO: A Revolutionary Approach 🆕
JSON-based Function Definitions: Function details are stored in external JSON files, enabling dynamic module loading.

Loose Coupling & Microservices: Each function can be developed, tested, and deployed as an independent module.

Flexible Scalability: Add new features by simply updating the JSON and module files, without modifying the core system.

JSON Example:
[
{
"name": "analyze_sentiment",
"module_path": "nlp_tools",
"func_name_in_module": "sentiment_analysis",
"example_usage": "analyze_sentiment(text=\"I love this product!\")"
}
]

Why MCO? 💡
Enhanced Development Efficiency: Developers can focus on their own modules with independent testing and deployment.

Simplified Error Management: Errors remain confined within their modules, enabling quick hotfixes.

Future-Proofing: With potential features like remote function calls (RPC), access control, auto-documentation, and a function marketplace, MCO paves the way for rapid innovation.

Practical Use & Community 🤝
The MCO implementation has been successfully tested on Vidraft’s LLM (based on Google Gemma-3)

reacted to nyuuzyou's post with 👍 2 days ago

Post

5087

🇷🇺 Russian Forum Messages Dataset - nyuuzyou/ruforum

Collection of approximately 58 million Russian forum messages featuring:

- Complete message content from Russian online forums spanning 2010-2025
- Comprehensive metadata including unique message IDs and timestamps
- Full text content preserving original user discussions and interactions
- Monolingual dataset focused exclusively on Russian language content

This dataset offers a unique textual archive of Russian online conversations suitable for text generation, sentiment analysis, and language modeling research. Released to the public domain under CC0 1.0 license.

reacted to katsukiai's post with 🧠 2 days ago

Post

3951

DeepFocus datasets are not allowed to be used in cases where mean is used in that dataset

Why?
├── This discussion is comments by the user. https://huggingface.co/JLouisBiz
├── Hello,
├── As a fork of a DeepSeek, you are required to give credit to DeepSeek according to the original MIT license. Could you please look into licensing terms and comply please?
├── I also do not see why are you making your own license, why don't you simple leave it with original MIT license?
└── I see that your license is also free software, but it brings legal problems when you are changing license, you are free to sublicense MIT licensed software, but re-licensing it without complying to initial terms is not allowed.
Unlicensed
├── DeepFocus
├── Wrong license and using modified license (Unpaper provided @aide-julius)
└── The dataset with the modified license does not use the same license as DeepSeek is using, EOS this license
Symbol
└── EOS
    └── End of service

Thank you,
Best Regards,
Sam from The KatsukiAI Team

2 replies

·

John Smith PRO

AI & ML interests

Recent Activity

Organizations

John6666's activity