Today's Fine-tuning & Training: Fastest-Growing Projects — September 23, 2026
Today's Fine-tuning & Training category highlights a diverse array of projects that cater to various aspects of AI model development and optimization. From decision-making models to dataset curators, these repositories are gaining traction as developers seek more efficient ways to fine-tune their models. Mapika/decider stands out with its high growth score, indicating significant interest in one-pass typed decisions with calibrated probabilities.
Mapika/decider is a project that fine-tunes from Qwen3.5-2B and offers a System One style model for making decisions based on calibrated probabilities. Its impressive growth score of 54.36 and 311 stars suggest strong community interest in its unique approach to decision-making models.
AkashPriyadarshii/jev-curate focuses on high-throughput synthetic and pretraining dataset sifting, utilizing Rust streaming core technology for efficient data processing. With a growth score of 26.10 and 27 stars, the project's active development and focus on performance optimization are likely driving its steady growth.
davefano/omarchy-trackpad-plus provides fine-grained trackpad controls and pointer-feel tuning for Omarchy devices. Despite being slightly less prominent with a growth score of 21.59 and 124 stars, the project's detailed approach to enhancing user experience through precise device control is attracting attention.
OpenMOSS/EasyWAM offers a unified framework for training, fine-tuning, and evaluating World Action Models. With a solid growth score of 14.95 and 312 stars, EasyWAM's comprehensive approach to managing AI models across different environments is clearly resonating with developers.
friday-memory/friday introduces an open-source persistent cognitive memory layer for AI coding agents, aiming to prevent amnesia in tools like Claude & Copilot. Its growth score of 13.07 and 54 stars indicate a growing need for persistent state management in AI applications.
njgymb/diffusers-sculptor presents an efficient framework for Stable Diffusion Fine-Tuning with a lightweight, config-driven pipeline. With a growth score of 11.31 and 56 stars, the project's focus on streamlined model training processes is attracting developers looking to optimize their workflows.
R0650N/spirit-flux-refiner offers an ultimate Flux1-LoRA Training Toolkit designed for AI creators seeking smart simplicity in their fine-tuning efforts. The toolkit's growth score of 11.29 and 55 stars reflect a strong community interest in its user-friendly approach to complex training tasks.
nrodriguez1997/lora-docker-aliyun-pipeline provides automated LoRA Training Docker builds for Alibaba Cloud, streamlining the process for developers working with this cloud platform. With the same growth score of 11.29 and 55 stars, its targeted approach to cloud-based training solutions is contributing to its popularity.
KingHsp/vram-sage-training presents a VRAM-smart fine-tuning suite optimized for SDXL and ANIMA models on 12GB GPUs. Its growth score of 11.29 and 55 stars indicate a strong demand for efficient model tuning in resource-constrained environments, highlighting the importance of optimizing GPU usage.
alrisqi/MrD-Optimizer-Lab showcases AI training and model optimization tools designed to enhance deep learning processes. With a consistent growth score of 11.29 and 55 stars, MrD's comprehensive suite is gaining traction among developers looking for robust tools to refine their models.
These projects collectively demonstrate the dynamic nature of the fine-tuning & training space, with active development in areas ranging from decision-making frameworks to efficient model optimization techniques.
Mapika/decider is a project that fine-tunes from Qwen3.5-2B and offers a System One style model for making decisions based on calibrated probabilities. Its impressive growth score of 54.36 and 311 stars suggest strong community interest in its unique approach to decision-making models.
AkashPriyadarshii/jev-curate focuses on high-throughput synthetic and pretraining dataset sifting, utilizing Rust streaming core technology for efficient data processing. With a growth score of 26.10 and 27 stars, the project's active development and focus on performance optimization are likely driving its steady growth.
davefano/omarchy-trackpad-plus provides fine-grained trackpad controls and pointer-feel tuning for Omarchy devices. Despite being slightly less prominent with a growth score of 21.59 and 124 stars, the project's detailed approach to enhancing user experience through precise device control is attracting attention.
OpenMOSS/EasyWAM offers a unified framework for training, fine-tuning, and evaluating World Action Models. With a solid growth score of 14.95 and 312 stars, EasyWAM's comprehensive approach to managing AI models across different environments is clearly resonating with developers.
friday-memory/friday introduces an open-source persistent cognitive memory layer for AI coding agents, aiming to prevent amnesia in tools like Claude & Copilot. Its growth score of 13.07 and 54 stars indicate a growing need for persistent state management in AI applications.
njgymb/diffusers-sculptor presents an efficient framework for Stable Diffusion Fine-Tuning with a lightweight, config-driven pipeline. With a growth score of 11.31 and 56 stars, the project's focus on streamlined model training processes is attracting developers looking to optimize their workflows.
R0650N/spirit-flux-refiner offers an ultimate Flux1-LoRA Training Toolkit designed for AI creators seeking smart simplicity in their fine-tuning efforts. The toolkit's growth score of 11.29 and 55 stars reflect a strong community interest in its user-friendly approach to complex training tasks.
nrodriguez1997/lora-docker-aliyun-pipeline provides automated LoRA Training Docker builds for Alibaba Cloud, streamlining the process for developers working with this cloud platform. With the same growth score of 11.29 and 55 stars, its targeted approach to cloud-based training solutions is contributing to its popularity.
KingHsp/vram-sage-training presents a VRAM-smart fine-tuning suite optimized for SDXL and ANIMA models on 12GB GPUs. Its growth score of 11.29 and 55 stars indicate a strong demand for efficient model tuning in resource-constrained environments, highlighting the importance of optimizing GPU usage.
alrisqi/MrD-Optimizer-Lab showcases AI training and model optimization tools designed to enhance deep learning processes. With a consistent growth score of 11.29 and 55 stars, MrD's comprehensive suite is gaining traction among developers looking for robust tools to refine their models.
These projects collectively demonstrate the dynamic nature of the fine-tuning & training space, with active development in areas ranging from decision-making frameworks to efficient model optimization techniques.