Today's Fine-tuning & Training: Fastest-Growing Projects — September 22, 2026
Today's Fine-tuning & Training category on GitHub continues to see a robust community engagement and development activity, with several projects emerging as leaders in their respective niches. The Mapika/decider project stands out with its innovative approach to one-pass decision-making models fine-tuned from Qwen3.5-2B, attracting significant attention among developers looking for advanced probabilistic decision systems.
Mapika/decider is a project that implements one-pass typed decisions using calibrated probabilities, which makes it particularly useful in applications requiring quick and accurate decision-making processes. With a growth score of 52.00 and 273 stars, the project's popularity suggests strong interest from developers working with probabilistic models and decision systems.
davefano/omarchy-trackpad-plus offers fine-grained control over trackpad settings, enhancing user experience through customized pointer-feel tuning for Omarchy devices. This tool has garnered a growth score of 23.75 and 124 stars, reflecting its appeal to developers interested in improving device-specific input controls.
OpenMOSS/EasyWAM provides a unified framework designed to streamline the process of training, fine-tuning, and evaluating World Action Models (WAMs). With a growth score of 15.44 and 309 stars, this project demonstrates its utility for researchers and developers looking to efficiently manage complex WAM tasks.
friday-memory/friday introduces an open-source persistent cognitive memory layer aimed at enhancing the functionality of AI coding agents such as Claude and Copilot by ensuring they retain context across sessions. Despite having fewer stars (51) compared to some other projects, its growth score of 15.00 indicates steady interest from developers seeking to improve the continuity and efficiency of their code generation tools.
njgymb/diffusers-sculptor is an efficient framework for fine-tuning Stable Diffusion models with a lightweight config-driven pipeline approach. The project's significant development activity, as indicated by its 100 commits in the last month, coupled with its growth score of 11.71 and 56 stars, highlights its relevance to developers working on AI art generation projects.
R0650N/spirit-flux-refiner presents a toolkit designed for smart fine-tuning of Flux1-LoRA models, emphasizing simplicity and ease-of-use for AI creators. With the same growth score (11.70) and stars count (55), this project's development intensity, measured by 100 commits in 30 days, suggests a growing community interested in optimizing LoRA training processes.
nrodriguez1997/lora-docker-aliyun-pipeline is an automated pipeline for building Docker images tailored for Alibaba Cloud, specifically for LoRA model training. The project's consistent development activity and steady growth score of 11.70 with 55 stars indicate its usefulness in the cloud-based AI model training ecosystem.
KingHsp/vram-sage-training offers a smart fine-tuning suite optimized for SDXL and ANIMA models on GPUs with limited VRAM, addressing the needs of developers working under resource constraints. Its growth score of 11.70 and 55 stars reflect its relevance to those interested in efficient model training solutions.
alrisqi/MrD-Optimizer-Lab provides a comprehensive suite for AI training and model optimization, offering tools that streamline deep learning workflows. With the same metrics as several other projects—growth score of 11.70 and 55 stars—the project's active development cycle is notable, indicating ongoing interest in its utility.
Leb947/modernbert-multi-task-studio rounds out Today's list with a modern framework for fine-tuning BERT models on multi-task learning tasks. Its growth score of 11.70 and 55 stars suggest steady engagement from developers looking to enhance their multi-modal AI applications through streamlined workflows.
Each project in the Fine-tuning & Training category offers unique solutions tailored to specific needs within the broader field of machine learning, highlighting a dynamic and evolving landscape where continuous innovation is key.
Mapika/decider is a project that implements one-pass typed decisions using calibrated probabilities, which makes it particularly useful in applications requiring quick and accurate decision-making processes. With a growth score of 52.00 and 273 stars, the project's popularity suggests strong interest from developers working with probabilistic models and decision systems.
davefano/omarchy-trackpad-plus offers fine-grained control over trackpad settings, enhancing user experience through customized pointer-feel tuning for Omarchy devices. This tool has garnered a growth score of 23.75 and 124 stars, reflecting its appeal to developers interested in improving device-specific input controls.
OpenMOSS/EasyWAM provides a unified framework designed to streamline the process of training, fine-tuning, and evaluating World Action Models (WAMs). With a growth score of 15.44 and 309 stars, this project demonstrates its utility for researchers and developers looking to efficiently manage complex WAM tasks.
friday-memory/friday introduces an open-source persistent cognitive memory layer aimed at enhancing the functionality of AI coding agents such as Claude and Copilot by ensuring they retain context across sessions. Despite having fewer stars (51) compared to some other projects, its growth score of 15.00 indicates steady interest from developers seeking to improve the continuity and efficiency of their code generation tools.
njgymb/diffusers-sculptor is an efficient framework for fine-tuning Stable Diffusion models with a lightweight config-driven pipeline approach. The project's significant development activity, as indicated by its 100 commits in the last month, coupled with its growth score of 11.71 and 56 stars, highlights its relevance to developers working on AI art generation projects.
R0650N/spirit-flux-refiner presents a toolkit designed for smart fine-tuning of Flux1-LoRA models, emphasizing simplicity and ease-of-use for AI creators. With the same growth score (11.70) and stars count (55), this project's development intensity, measured by 100 commits in 30 days, suggests a growing community interested in optimizing LoRA training processes.
nrodriguez1997/lora-docker-aliyun-pipeline is an automated pipeline for building Docker images tailored for Alibaba Cloud, specifically for LoRA model training. The project's consistent development activity and steady growth score of 11.70 with 55 stars indicate its usefulness in the cloud-based AI model training ecosystem.
KingHsp/vram-sage-training offers a smart fine-tuning suite optimized for SDXL and ANIMA models on GPUs with limited VRAM, addressing the needs of developers working under resource constraints. Its growth score of 11.70 and 55 stars reflect its relevance to those interested in efficient model training solutions.
alrisqi/MrD-Optimizer-Lab provides a comprehensive suite for AI training and model optimization, offering tools that streamline deep learning workflows. With the same metrics as several other projects—growth score of 11.70 and 55 stars—the project's active development cycle is notable, indicating ongoing interest in its utility.
Leb947/modernbert-multi-task-studio rounds out Today's list with a modern framework for fine-tuning BERT models on multi-task learning tasks. Its growth score of 11.70 and 55 stars suggest steady engagement from developers looking to enhance their multi-modal AI applications through streamlined workflows.
Each project in the Fine-tuning & Training category offers unique solutions tailored to specific needs within the broader field of machine learning, highlighting a dynamic and evolving landscape where continuous innovation is key.