Today's Fine-tuning & Training: Fastest-Growing Projects — October 13, 2026
Today's Fine-tuning & Training category showcases a mix of innovative projects catering to various domains such as cybersecurity, decision-making models, and multimodal training. The standout project, Mapika/decider, continues to gain traction for its unique approach to making decisions with calibrated probabilities using System One style modeling. Additionally, several tools like AcademiaSD's LoRA trainer and Liuziyu77's Valen demonstrate the growing interest in fine-tuning large language models (LLMs) for specialized applications.
Mapika/decider is a one-pass typed decision system that leverages calibrated probabilities based on the System One style model, fine-tuned from Qwen3.5-2B. Its growth score of 34.69 and an impressive 1,132 stars highlight its relevance in probabilistic decision-making frameworks.
JoasASantos/Offensive-Security-AI-Models offers uncensored AI models or those fine-tuned specifically for cybersecurity tasks, attracting a substantial following with 674 stars despite a relatively lower growth score of 29.69. This project's focus on specialized security applications likely contributes to its steady growth.
Liuziyu77/Valen enables users to train Jev-like multimodal models, incorporating vision capabilities into the System One model framework. With a growth score of 29.43 and 769 stars, Valen stands out for providing a versatile platform for advanced multimodal training tasks.
AcademiaSD/AcademiaSD_LoRAlab-TrainerStudio facilitates LoRA training across various models including Qwen-Image 2.1 and SDXL on NVIDIA GPUs with limited VRAM capacity. Its growth score of 23.33 and 73 stars indicate its utility in optimizing model training for resource-constrained environments.
Empero-org/brewery-ai provides an intuitive console agent designed to simplify the fine-tuning process, requiring minimal experience from users. With a growth score of 18.78 and 206 stars, this tool addresses the gap between AI model complexity and ease of use in fine-tuning workflows.
SimpleJev/JevAny offers open infrastructure for training, evaluating, and deploying System 1 decision models across different language and multimodal backbones. Its growth score of 16.83 and 77 stars suggest a growing interest in standardized frameworks for model evaluation and deployment.
Veda-sparse/Miowtion focuses on sparse acceleration and LoRA fine-tuning of MiniMax-H3, optimized for resource-constrained machines. With a growth score of 16.25 and 38 stars, this project demonstrates the demand for efficient training solutions in limited hardware environments.
Mehmetbulutaktas458-dev/Aozora-Forge is a toolkit designed to train full-quality AI models on GPUs with as little as 12GB VRAM. Its growth score of 14.93 and 57 stars reflect the increasing need for tools that can handle large model training efficiently within resource constraints.
09854145774/Bert-Domain-Shift-Lab guides users through domain adaptation techniques using BERT models from Hugging Face, aiming to fine-tune pretrained models effectively. Its growth score of 14.91 and 56 stars indicate its growing importance in the domain-specific training and adaptation space.
These projects collectively illustrate the dynamic landscape of AI model fine-tuning and training, with a range of tools addressing specific needs from decision-making frameworks to security applications and efficient resource management.
Mapika/decider is a one-pass typed decision system that leverages calibrated probabilities based on the System One style model, fine-tuned from Qwen3.5-2B. Its growth score of 34.69 and an impressive 1,132 stars highlight its relevance in probabilistic decision-making frameworks.
JoasASantos/Offensive-Security-AI-Models offers uncensored AI models or those fine-tuned specifically for cybersecurity tasks, attracting a substantial following with 674 stars despite a relatively lower growth score of 29.69. This project's focus on specialized security applications likely contributes to its steady growth.
Liuziyu77/Valen enables users to train Jev-like multimodal models, incorporating vision capabilities into the System One model framework. With a growth score of 29.43 and 769 stars, Valen stands out for providing a versatile platform for advanced multimodal training tasks.
AcademiaSD/AcademiaSD_LoRAlab-TrainerStudio facilitates LoRA training across various models including Qwen-Image 2.1 and SDXL on NVIDIA GPUs with limited VRAM capacity. Its growth score of 23.33 and 73 stars indicate its utility in optimizing model training for resource-constrained environments.
Empero-org/brewery-ai provides an intuitive console agent designed to simplify the fine-tuning process, requiring minimal experience from users. With a growth score of 18.78 and 206 stars, this tool addresses the gap between AI model complexity and ease of use in fine-tuning workflows.
SimpleJev/JevAny offers open infrastructure for training, evaluating, and deploying System 1 decision models across different language and multimodal backbones. Its growth score of 16.83 and 77 stars suggest a growing interest in standardized frameworks for model evaluation and deployment.
Veda-sparse/Miowtion focuses on sparse acceleration and LoRA fine-tuning of MiniMax-H3, optimized for resource-constrained machines. With a growth score of 16.25 and 38 stars, this project demonstrates the demand for efficient training solutions in limited hardware environments.
Mehmetbulutaktas458-dev/Aozora-Forge is a toolkit designed to train full-quality AI models on GPUs with as little as 12GB VRAM. Its growth score of 14.93 and 57 stars reflect the increasing need for tools that can handle large model training efficiently within resource constraints.
09854145774/Bert-Domain-Shift-Lab guides users through domain adaptation techniques using BERT models from Hugging Face, aiming to fine-tune pretrained models effectively. Its growth score of 14.91 and 56 stars indicate its growing importance in the domain-specific training and adaptation space.
These projects collectively illustrate the dynamic landscape of AI model fine-tuning and training, with a range of tools addressing specific needs from decision-making frameworks to security applications and efficient resource management.