Today's Fine-tuning & Training: Fastest-Growing Projects — May 02, 2026
Today's the Fine-tuning & Training space, we saw a surge of interest in tools that enable efficient and effective fine-tuning of large language models (LLMs) and other AI architectures. Many of the top-growing repositories focus on multimodal fine-tuning, LoRA training, and innovative approaches to improving model performance without increasing inference costs.
mattmireles/gemma-tuner-multimodal takes the top spot with a growth score of 46.12 and 1,400 stars. This repository provides a framework for fine-tuning Gemma 4 and 3n models with audio, images, and text on Apple Silicon using PyTorch and Metal Performance Shaders. Its popularity can be attributed to its ability to efficiently fine-tune models across multiple modalities.
QingGo/engram-peft comes in second with a growth score of 16.15 and 31 stars. This unofficial implementation of DeepSeek Engram enables the injection of high-capacity conditional memory into LLMs via sparse retrieval PEFT without increasing inference FLOPs. Its growing popularity can be attributed to its innovative approach to improving model performance.
UNfukashigi/Anima-LoRA-Factory has a growth score of 11.42 and 24 stars. This user-friendly GUI tool is designed for training LoRAs for the next-generation Anima diffusion models, making it an attractive option for researchers and developers looking to fine-tune their models efficiently.
ZJU-OmniAI/GFT boasts a growth score of 9.53 and 29 stars. This repository provides a novel approach to reward fine-tuning with unbiased group advantages and dynamic coefficient rectification, which has sparked interest in the community.
semidark/kokoro-deutsch has a growth score of 5.68 and 31 stars. This project provides a complete training recipe for fine-tuning Kokoro-82M on a new language, specifically German, making it a valuable resource for those looking to adapt models to new languages.
Dynamis-Labs/spectralquant has a growth score of 3.63 and an impressive 130 stars. Although its growth rate is slower than others, this repository's unique approach to breaking TurboQuant's compression limit via spectral structure has garnered significant attention from the community.
Mintzs/oogaboogalm has a growth score of 2.75 and 46 stars. This project explores the idea of baking caveman system prompts and skills into AI models through fine-tuning, which has piqued the interest of researchers looking to optimize model performance.
Goekdeniz-Guelmez/moshi-finetune-mlx rounds out our list with a growth score of 0.64 and 25 stars. This repository provides a tool for fine-tuning Moshi (Native, Real-Time, Speech-to-Speech) models on Apple Silicon, making it an attractive option for those working with speech-to-speech applications.
Overall, Today's trends in Fine-tuning & Training highlight the community's focus on efficient and effective methods for improving model performance across various architectures and modalities.
mattmireles/gemma-tuner-multimodal takes the top spot with a growth score of 46.12 and 1,400 stars. This repository provides a framework for fine-tuning Gemma 4 and 3n models with audio, images, and text on Apple Silicon using PyTorch and Metal Performance Shaders. Its popularity can be attributed to its ability to efficiently fine-tune models across multiple modalities.
QingGo/engram-peft comes in second with a growth score of 16.15 and 31 stars. This unofficial implementation of DeepSeek Engram enables the injection of high-capacity conditional memory into LLMs via sparse retrieval PEFT without increasing inference FLOPs. Its growing popularity can be attributed to its innovative approach to improving model performance.
UNfukashigi/Anima-LoRA-Factory has a growth score of 11.42 and 24 stars. This user-friendly GUI tool is designed for training LoRAs for the next-generation Anima diffusion models, making it an attractive option for researchers and developers looking to fine-tune their models efficiently.
ZJU-OmniAI/GFT boasts a growth score of 9.53 and 29 stars. This repository provides a novel approach to reward fine-tuning with unbiased group advantages and dynamic coefficient rectification, which has sparked interest in the community.
semidark/kokoro-deutsch has a growth score of 5.68 and 31 stars. This project provides a complete training recipe for fine-tuning Kokoro-82M on a new language, specifically German, making it a valuable resource for those looking to adapt models to new languages.
Dynamis-Labs/spectralquant has a growth score of 3.63 and an impressive 130 stars. Although its growth rate is slower than others, this repository's unique approach to breaking TurboQuant's compression limit via spectral structure has garnered significant attention from the community.
Mintzs/oogaboogalm has a growth score of 2.75 and 46 stars. This project explores the idea of baking caveman system prompts and skills into AI models through fine-tuning, which has piqued the interest of researchers looking to optimize model performance.
Goekdeniz-Guelmez/moshi-finetune-mlx rounds out our list with a growth score of 0.64 and 25 stars. This repository provides a tool for fine-tuning Moshi (Native, Real-Time, Speech-to-Speech) models on Apple Silicon, making it an attractive option for those working with speech-to-speech applications.
Overall, Today's trends in Fine-tuning & Training highlight the community's focus on efficient and effective methods for improving model performance across various architectures and modalities.