Dcas89 (Dcas89)

reacted to sergiopaniego's post with 🤗 about 1 month ago

Post

2237

ICYMI, transformers v5 is out!

Grab a coffee ☕ and go read the announcement blog https://huggingface.co/blog/transformers-v5

reacted to Jofthomas's post with 🔥 about 1 month ago

Post

3582

The new Mistral 3 models are here !

Today, we announce Mistral 3, the next generation of Mistral models. Mistral 3 includes three state-of-the-art small, dense models (14B, 8B, and 3B) and Mistral Large 3 – our most capable model to date – a sparse mixture-of-experts trained with 41B active and 675B total parameters.

All models are released under the Apache 2.0 license.

Ministrals :
https://huggingface.co/collections/mistralai/ministral-3

Mistral Large 3:
https://huggingface.co/collections/mistralai/mistral-large-3

2 replies

·

reacted to Kseniase's post with 🔥 2 months ago

Post

11161

11 Fascinating new Policy Optimization techniques

Policy optimization (PO) algorithms are central to training AI models with preference-based feedback. In recent weeks, numerous new PO methods have emerged that build on or replace the popular PPO and GRPO, solving their issues. Here are 11 of them:

1. BAlanced Policy Optimization (BAPO) → BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clipping (2510.18927)
Dynamically adjusting the clipping bounds in PPO-style updates to balance positive and negative gradients and prevent entropy collapse

2. Training-Free GRPO → Training-Free Group Relative Policy Optimization (2510.08191)
Instead of using numeric rewards, it compares rollouts semantically to distill useful knowledge as a token prior, which is then applied during inference to guide the model’s behavior

3. Asymmetric Importance Sampling Policy Optimization (ASPO) → ASPO: Asymmetric Importance Sampling Policy Optimization (2510.06062)
Fixes imbalanced token weighting in LLM training. It flips the importance sampling ratios for positive tokens to correct over- and under-updates, and adds a soft dual-clipping step to keep gradients stable

4. In-Context Steered Policy Optimization (ICPO) → https://arxiv.org/abs/2510.26519
Uses a model’s own in-context learning ability to guide training with existing data. It combines Mixed-Policy GRPO with Implicit Expert Forcing to expand exploration and adds Expert Region Reject Sampling and Annealed Expert-Bonus Reward Shaping to ensure stability and balanced expert influence

5. Graph-Enhanced Policy Optimization (GEPO) → https://arxiv.org/abs/2510.26270
Builds a graph of an agent’s experiences to understand how different states connect, guide exploration and assign rewards more effectively

6. Information Gain-based Policy Optimization (IGPO) → Information Gain-based Policy Optimization: A Simple and Effective Approach for Multi-Turn LLM Agents (2510.14967)
Uses the model’s own belief updates to create dense, informative feedback for smoother multi-turn learning

Read further below ⬇️
If you like this, also subscribe to the Turing post: https://www.turingpost.com/subscribe

2 replies

·

reacted to prithivMLmods's post with 👍 3 months ago

Post

4554

Try the Hugging Face Space demo for Logics-MLLM/Logics-Parsing, the latest multimodal VLM from the Logics Team at Alibaba Group. It enables end-to-end document parsing with precise content extraction in markdown format, and it also generates a clean HTML representation of the document while preserving its logical structure. 🤗🔥

Additionally, I’ve integrated one of my recent works — prithivMLmods/Gliese-OCR-7B-Post1.0 — which also excels at document comprehension.

⭐ Space / App : prithivMLmods/VLM-Parsing
📄 Technical Report by the Logics Team, Alibaba Group : Logics-Parsing Technical Report (2509.19760)
🖖 MM: VLM-Parsing: prithivMLmods/mm-vlm-parsing-68e33e52bfb9ae60b50602dc
⚡ Collections : prithivMLmods/multimodal-implementations-67c9982ea04b39f0608badb0

Other Pages:

➔ Multimodal VLMs - July'25 : prithivMLmods/multimodal-vlms-until-july25-688312e6b840e1e156f13027
➔ Multimodal VLMs - Aug'25 : prithivMLmods/multimodal-vlms-aug25-68a56aac39fe8084f3c168bd
➔ VL caption — < Sep 15 ’25 : prithivMLmods/vl-caption-sep-15-25-68c7f6d737985c63c13e2391

.
.
.
To know more about it, visit the app page or the respective model page!!

reacted to MonsterMMORPG's post with 🔥 3 months ago

Post

3301

Ovi - Generate Videos With Audio Like VEO 3 or SORA 2 - Run Locally - Open Source for Free

Download and install : https://www.patreon.com/posts/140393220

Quick demo tutorial : https://youtu.be/uE0QabiHmRw

Ovi: Twin Backbone Cross-Modal Fusion for Audio-Video Generation

Project page : https://aaxwaz.github.io/Ovi/

SECourses Ovi Pro Premium App Features

Full scale ultra advanced app for Ovi - an open source project that can generate videos from both text prompts and image + text prompts with real audio.

Project page is here : https://aaxwaz.github.io/Ovi/

I have developed an ultra advanced Gradio app and much better pipeline that fully supports block swapping

Now we can generate full quality videos with as low as 8.2 GB VRAM

Hopefully I will work on dynamic on load FP8_Scaled tomorrow to improve VRAM even further

So more VRAM optimizations will come hopefully tomorrow

Our implemented block swapping is the very best one out there - I took the approach from famous Kohya Musubi tuner

The 1-click installer will install into Python 3.10.11 venv and will auto download models as well so it is literally 1-click

My installer auto installs with Torch 2.8, CUDA 12.9, Flash Attention 2.8.3 and it supports literally all GPUs like RTX 3000 series, 4000 series, 5000 series, H100, B200, etc

All generations will be saved inside outputs folder and we support so many features like batch folder processing, number of generations, full preset save and load

This is a rush release (in less than a day) so there can be errors please let me know and I will hopefully improve the app

Look the examples to understand how to prompt the model that is extremely important

RTX 5090 can run it without any block swap with just cpu-offloading - really fast

2 replies

·

reacted to sergiopaniego's post with 🔥 4 months ago

Post

1410

This summer TRL leveled up for multimodal alignment 🌞

✅ New VLM alignment methods (MPO, GRPO, GSPO)
✅ Extended RLOO & Online DPO for VLMs
✅ Native SFT support
✅ Ready-to-use training scripts

🔗 https://huggingface.co/blog/trl-vlm-alignment

reacted to vikhyatk's post with 🔥 4 months ago

Post

4909

Just released a preview of Moondream 3! moondream/moondream3-preview

This is a 9B parameter, 2B active MoE VLM with state of the art visual reasoning capabilities.

More details in the release blog post: https://moondream.ai/blog/moondream-3-preview

3 replies

·

reacted to sergiopaniego's post with 🤗 5 months ago

Post

2932

So you can now SFT a model with hf jobs + TRL in ONE command lol 🏎️💨

Without worrying about infrastructure since it runs entirely on HF!

docs: https://huggingface.co/docs/huggingface_hub/main/en/guides/jobs
blog: https://huggingface.co/blog/hf-cli

reacted to prithivMLmods's post with 👍 5 months ago

Post

4412

On the verge of releasing Poseidon-Reasoning-5M, a dataset built to excel in general thought processes, mathematics, and science across a diverse mixture of domains, I’m also dropping the Gargantua-R1-Compact dataset, a collection of over six million high-quality reasoning QA pair traces. 🤗🚀

✦ Gargantua-R1-Compact : prithivMLmods/Gargantua-R1-Compact

from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Gargantua-R1-Compact", split="train")

Additionally, I’m adding the mini version of Gargantua — the Gargantua-R1-Wee : prithivMLmods/Gargantua-R1-Wee

from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Gargantua-R1-Wee", split="train")

The composition spans 73.93% core mathematical reasoning involving problems, proofs, and computational challenges, 12.11% across diverse scientific domains such as physics, chemistry, biology, and interdisciplinary topics, 11.35% in competitive coding covering algorithms and data structures, 1.37% in academic science focusing on research-level methodology, 0.95% in creative and analytical reasoning through logic puzzles and problem-solving tasks, 0.25% in specialized technical areas like MLOps, LLMs, diffusion models, and CUDA, and 0.06% involving data from graphs and charts converted into structured JSON formats. Designed with both rich contextual depth and formal structural clarity, Gargantua-R1-Compact is an optimal resource for advancing research in symbolic reasoning, interpretability, and high-precision question answering in mathematical domains.

✦ Collection : prithivMLmods/gargantua-r1-mod-6896bfd7834e82b89ad2b38b

To know more about it, visit the dataset card of the respective dataset. !!

reacted to codelion's post with 🔥 5 months ago

Post

4716

Released 17 production-ready adaptive text classifiers that learn from just 100 examples per class and continuously improve without retraining.

These models achieve 93% average accuracy across enterprise use cases like email routing, fraud detection, document classification, and support ticket categorization. Built on ModernBERT with prototype memory and elastic weight consolidation.

Key benefits: 90% cost reduction vs API solutions, 90-120ms local inference, dynamic class addition, and zero vendor lock-in.

All models available under adaptive-classifier organization. Install with pip install adaptive-classifier.

Full technical details: https://huggingface.co/blog/codelion/enterprise-ready-classifiers
Code: https://github.com/codelion/adaptive-classifier

2 replies

·

reacted to sergiopaniego's post with 🤗 6 months ago

Post

2748

We just released TRL v0.20 with major multimodal upgrades!

👁️ VLM support for GRPO (highly requested by the community!)
🎞️ New GSPO trainer (from @Qwen , released last week, VLM-ready)
🐙 New MPO trainer (multimodal by design, as in the paper)

📝 Full release notes here: https://github.com/huggingface/trl/releases/tag/v0.20.0

reacted to MaziyarPanahi's post with 🔥 6 months ago

Post

12977

🧬 Breaking news in Clinical AI: Introducing the OpenMed NER Model Discovery App on Hugging Face 🔬

OpenMed is back! 🔥 Finding the right biomedical NER model just became as precise as a PCR assay!

I'm thrilled to unveil my comprehensive OpenMed Named Entity Recognition Model Discovery App that puts 384 specialized biomedical AI models at your fingertips.

🎯 Why This Matters in Healthcare AI:
Traditional clinical text mining required hours of manual model evaluation. My Discovery App instantly connects researchers, clinicians, and data scientists with the exact NER models they need for their biomedical entity extraction tasks.

🔬 What You Can Discover:
✅ Pharmacological Models - Extract "chemical compounds", "drug interactions", and "pharmaceutical" entities from clinical notes
✅ Genomics & Proteomics - Identify "DNA sequences", "RNA transcripts", "gene variants", "protein complexes", and "cell lines"
✅ Pathology & Disease Detection - Recognize "pathological formations", "cancer types", and "disease entities" in medical literature
✅ Anatomical Recognition - Map "anatomical systems", "tissue types", "organ structures", and "cellular components"
✅ Clinical Entity Extraction - Detect "organism species", "amino acids", 'protein families", and "multi-tissue structures"

💡 Advanced Features:
🔍 Intelligent Entity Search - Find models by specific biomedical entities (e.g., "Show me models detecting CHEM + DNA + Protein")
🏥 Domain-Specific Filtering - Browse by Oncology, Pharmacology, Genomics, Pathology, Hematology, and more
📊 Model Architecture Insights - Compare BERT, RoBERTa, and DeBERTa implementations
⚡ Real-Time Search - Auto-filtering as you type, no search buttons needed
🎨 Clinical-Grade UI - Beautiful, intuitive interface designed for medical professionals

Ready to revolutionize your biomedical NLP pipeline?

🔗 Try it now: OpenMed/openmed-ner-models
🧬 Built with: Gradio, Transformers, Advanced Entity Mapping

5 replies

·

reacted to prithivMLmods's post with 🤗 6 months ago

Post

5160

Upgraded the step-by-step notebook for fine-tuning SigLIP2 on domain-specific image classification tasks. The notebook supports both datasets with predefined train/test splits and those with only a train split, making it suitable for low-resource, custom, and real-world classification scenarios. 📢👉

➺ FineTuning-SigLIP2-Notebook : prithivMLmods/FineTuning-SigLIP2-Notebook

➺ GitHub : https://github.com/PRITHIVSAKTHIUR/FineTuning-SigLIP-2

➺ In the first, datasets include predefined train and test splits, enabling conventional supervised learning and generalization evaluation : prithivMLmods/FineTuning-SigLIP2-Notebook (.ipynb)

➺ In the second scenario, only a training split is available; in such cases, the training set is either partially reserved for validation or reused entirely for evaluation : prithivMLmods/FineTuning-SigLIP2-Notebook (.ipynb)

This flexibility supports experimentation in constrained or domain-specific settings, where standard test annotations may not exist.

reacted to giadap's post with 🤗 6 months ago

Post

1272

🤖 Technology means power, and whoever owns the technology owns the power.

Thrilled to share insights from my recent interview with MIT Technology Review about the growing movement toward local LLMs and what it means for AI democratization. Read here: https://www.technologyreview.com/2025/07/17/1120391/how-to-run-an-llm-on-your-laptop/

🤔 Why this matters: When we use "free" online AI services, we're often the product. Our conversations become training data, our personal stories get "cooked into" models, and our privacy becomes a commodity. But there's an alternative path forward.

💡 The power shift is real: Local LLMs aren't just about privacy; they're about redistributing AI power away from a handful of tech giants. When individuals, organizations, and even entire nations can run their own models, we're democratizing access to AI capabilities.

🤗 At Hugging Face, we're proud to be at the center of this transformation. Our platform hosts the world's largest library of freely downloadable models, making cutting-edge AI accessible to everyone -- from researchers and developers to curious individuals who want to experiment on their laptops or even smartphones.

The technical barriers that once required $$$ server racks are crumbling. Today, anyone with basic computer skills can download a model, run it locally, and maintain complete control over their AI interactions. No sudden algorithm changes, no data harvesting, no corporate gatekeeping.

This is about technical convenience, but especially about technological sovereignty. When AI power is concentrated in a few hands, we risk creating new forms of digital dependency. Local models offer a path toward genuine AI literacy and independence.

🚀 The future of AI should be open, accessible, and in the hands of the many, not the few. What are your thoughts on AI democratization? Have you experimented with local models yet?

reacted to AdinaY's post with 🔥 6 months ago

Post

1523

M2-Reasoning🔥 a unified multimodal model for general (math, logic) and spatial (motion, physics, orientation) reasoning, released by AntGroup.

Model:
inclusionAI/M2-Reasoning
Paper:
M2-Reasoning: Empowering MLLMs with Unified General and Spatial Reasoning (2507.08306)

✨ 7B with MIT license
✨ 294K high quality samples via novel data pipeline
✨ Dynamic multi-task training to resolve task conflicts

reacted to sergiopaniego's post with 👍 6 months ago

Post

1698

🧑‍🍳 New Multimodal Fine-Tuning Recipe 🧑‍🍳

⚡️ In this new @huggingface Cookbook recipe, I walk you though the process of fine tuning a Visual Language Model (VLM) for Object Detection with Visual Grounding, using TRL.

🔍 Object detection typically involves detecting categories in images (e.g., vase).

By combining it with visual grounding, we add contextual understanding so instead of detecting just "vase", we can detect "middle vase" in an image.

VLMs are super powerful!

In this case, I use PaliGemma 2 which already supports object detection and extend it to also add visual grounding.

🤗 Check it out here: https://huggingface.co/learn/cookbook/fine_tuning_vlm_object_detection_grounding

reacted to prithivMLmods's post with 🤗 6 months ago

Post

4108

Dropping the general-purpose reasoning dataset Poseidon-Reasoning-5M, which supports general thought processes, math, and science — featuring a diverse mixture of domains 🌊 : prithivMLmods/Poseidon-Reasoning-5M

from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Poseidon-Reasoning-5M", split="data")

The compact version is as follows — Poseidon-Reasoning-Mini-300K : prithivMLmods/Poseidon-Reasoning-Mini-300K

from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Poseidon-Reasoning-Mini-300K", split="train")

Collection : prithivMLmods/poseidon-reasoning-6879ca98e118b307c781a9ba

reacted to Abhaykoul's post with 🔥 6 months ago

Post

3162

🎉 Dhanishtha-2.0-preview-0725 is Now Live

The Intermediate Thinking Model just got even better.
With the new update, Dhanishtha is now sharper, smarter, and trained further on tool use

🧠 What Makes Dhanishtha Different?
Unlike standard COT models that give one-shot responses, Dhanishtha thinks in layers:

> Think → Answer → Rethink → Improve → Rethink again if needed.

HelpingAI/Dhanishtha-2.0-preview-0725

reacted to hba123's post with 👀 6 months ago

Post

1376

As promised, and after the request of many, we have managed to fit in the first live session about Ark that we will be giving on the 28th of July.

pip install ark-robotics

For those who are already in the messaging channel, all is done, no need to do anything :-D For those interested in registering, please write to me at [email protected] - then I can add you and send you the invite.

We chose the timing to be 5 pm UK after consulting many of the interested people. Hope it works well for you too?

See you soon! Till then, have fun looking and using Ark: https://arkrobotics.notion.site/ARK-Home-22be053d9c6f8096bcdbefd6276aba61

1 reply

·

reacted to etemiz's post with 🚀 6 months ago

Post

535

Qwen 3 32B fine tuning with Unsloth is going well. It does not resist to faith training like Gemma 3 did. I may open weights at some point.

Qwen 3 is more capable than Gemma 3, and after fine tuning it will probably be more aligned. It does not get into "chanting" (repetition of words or sentences) even when temp = 0.

The base training by Qwen was done using 36T tokens on a 32B parameters. About 2 times bigger than Gemma 3's ratio and 4 times bigger than Llama 3's ratio. This is a neat model. My fine tuning is more like billions of tokens. We will see if billions is enough to "convince" trillions.

Dcas89 PRO

AI & ML interests

Recent Activity

Organizations

Dcas89 PRO

AI & ML interests

Recent Activity

Organizations

Dcas89's activity