Today's Fine-tuning & Training: Fastest-Growing Projects — April 27, 2026
Today's the Fine-tuning & Training space, we're seeing a surge of interest in multimodal models and innovative fine-tuning techniques. Several repositories are gaining traction by providing user-friendly tools for training LoRAs, fine-tuning large language models, and leveraging sparse retrieval methods. Meanwhile, researchers are exploring new approaches to improve model performance and efficiency.
mattmireles/gemma-tuner-multimodal is a standout repository this week, with a growth score of 57.27 and 1,391 stars. This project provides a PyTorch-based implementation for fine-tuning Gemma 4 and 3n models on Apple Silicon, using audio, images, and text inputs. Its popularity can be attributed to the growing interest in multimodal models and the increasing availability of powerful hardware.
QingGo/engram-peft has gained significant attention with a growth score of 21.53 and 31 stars. This repository offers an unofficial implementation of DeepSeek Engram, which injects high-capacity conditional memory into large language models via sparse retrieval PEFT without increasing inference FLOPs. Its growth is likely driven by researchers seeking to improve the efficiency and performance of their LLMs.
UNfukashigi/Anima-LoRA-Factory has a growth score of 17.75 and 23 stars, thanks to its user-friendly GUI tool for training LoRAs for next-generation Anima diffusion models. This repository's popularity stems from the growing interest in Anima models and the need for accessible tools for fine-tuning.
ZJU-OmniAI/GFT boasts a growth score of 13.86 and 29 stars, with its focus on reward fine-tuning using unbiased group advantages and dynamic coefficient rectification. Researchers are likely drawn to this repository due to its innovative approach to improving model performance.
SUM-INNOVATION/RUMUS has a growth score of 9.78 and 186 stars, offering a Rust-based framework for training neural networks. Its popularity can be attributed to the growing interest in alternative deep learning frameworks and the efficiency offered by Rust.
WillowHe/EvoOpt_oppangu_optimization_model has gained significant attention with a growth score of 9.55 and 514 stars. This repository provides solutions leveraging Openpangu - 7B as the base model for fine-tuning large language models in operations research optimization tasks. Its popularity stems from the growing interest in applying LLMs to real-world problems.
semidark/kokoro-deutsch has a growth score of 7.06 and 30 stars, offering a complete training recipe for fine-tuning Kokoro-82M on German language inputs. Researchers are likely drawn to this repository due to its focus on low-resource languages and the need for accessible fine-tuning recipes.
Dynamis-Labs/spectralquant boasts a growth score of 4.41 and 128 stars, with its innovative approach to breaking TurboQuant's compression limit via spectral structure. This repository's popularity can be attributed to the growing interest in improving model efficiency.
Mintzs/oogaboogalm has gained attention with a growth score of 3.53 and 45 stars, exploring the idea of baking caveman system prompts and skills into models through fine-tuning. Researchers are likely drawn to this repository due to its novel approach to reducing token usage.
PentesterFlow/OffensiveSET rounds out our list with a growth score of 3.24 and 72 stars, offering an MCP server for generating high-quality pentesting conversation datasets for LLM fine-tuning. This repository's popularity stems from the growing interest in security-focused applications of LLMs.
mattmireles/gemma-tuner-multimodal is a standout repository this week, with a growth score of 57.27 and 1,391 stars. This project provides a PyTorch-based implementation for fine-tuning Gemma 4 and 3n models on Apple Silicon, using audio, images, and text inputs. Its popularity can be attributed to the growing interest in multimodal models and the increasing availability of powerful hardware.
QingGo/engram-peft has gained significant attention with a growth score of 21.53 and 31 stars. This repository offers an unofficial implementation of DeepSeek Engram, which injects high-capacity conditional memory into large language models via sparse retrieval PEFT without increasing inference FLOPs. Its growth is likely driven by researchers seeking to improve the efficiency and performance of their LLMs.
UNfukashigi/Anima-LoRA-Factory has a growth score of 17.75 and 23 stars, thanks to its user-friendly GUI tool for training LoRAs for next-generation Anima diffusion models. This repository's popularity stems from the growing interest in Anima models and the need for accessible tools for fine-tuning.
ZJU-OmniAI/GFT boasts a growth score of 13.86 and 29 stars, with its focus on reward fine-tuning using unbiased group advantages and dynamic coefficient rectification. Researchers are likely drawn to this repository due to its innovative approach to improving model performance.
SUM-INNOVATION/RUMUS has a growth score of 9.78 and 186 stars, offering a Rust-based framework for training neural networks. Its popularity can be attributed to the growing interest in alternative deep learning frameworks and the efficiency offered by Rust.
WillowHe/EvoOpt_oppangu_optimization_model has gained significant attention with a growth score of 9.55 and 514 stars. This repository provides solutions leveraging Openpangu - 7B as the base model for fine-tuning large language models in operations research optimization tasks. Its popularity stems from the growing interest in applying LLMs to real-world problems.
semidark/kokoro-deutsch has a growth score of 7.06 and 30 stars, offering a complete training recipe for fine-tuning Kokoro-82M on German language inputs. Researchers are likely drawn to this repository due to its focus on low-resource languages and the need for accessible fine-tuning recipes.
Dynamis-Labs/spectralquant boasts a growth score of 4.41 and 128 stars, with its innovative approach to breaking TurboQuant's compression limit via spectral structure. This repository's popularity can be attributed to the growing interest in improving model efficiency.
Mintzs/oogaboogalm has gained attention with a growth score of 3.53 and 45 stars, exploring the idea of baking caveman system prompts and skills into models through fine-tuning. Researchers are likely drawn to this repository due to its novel approach to reducing token usage.
PentesterFlow/OffensiveSET rounds out our list with a growth score of 3.24 and 72 stars, offering an MCP server for generating high-quality pentesting conversation datasets for LLM fine-tuning. This repository's popularity stems from the growing interest in security-focused applications of LLMs.