Today's Fine-tuning & Training: Fastest-Growing Projects — April 28, 2026
The Fine-tuning & Training space has seen significant activity this week, with a focus on multimodal fine-tuning, parameter-efficient methods, and user-friendly tools for training large language models. Several repositories have gained traction by providing innovative solutions for improving model performance, reducing computational costs, or facilitating the development of specialized models.
mattmireles/gemma-tuner-multimodal has taken the top spot with a growth score of 54.55 and 1,391 stars, as it enables users to fine-tune Gemma 4 and 3n models on Apple Silicon using PyTorch and Metal Performance Shaders, making it an attractive option for those looking to leverage multimodal capabilities. The repository's high commit activity over the past month (100 commits) suggests a strong interest in this specific use case.
QingGo/engram-peft has gained significant attention with a growth score of 20.19 and 31 stars, as it provides an unofficial implementation of DeepSeek Engram, allowing users to inject high-capacity conditional memory into large language models via sparse retrieval PEFT without increasing inference FLOPs. The repository's popularity can be attributed to the growing interest in parameter-efficient methods for fine-tuning large language models.
UNfukashigi/Anima-LoRA-Factory has seen notable growth with a score of 15.78 and 23 stars, as it offers a user-friendly GUI tool designed specifically for training LoRAs for next-generation Anima diffusion models. The repository's popularity can be attributed to the increasing demand for specialized tools that simplify the process of fine-tuning large language models.
ZJU-OmniAI/GFT has gained traction with a growth score of 12.71 and 29 stars, as it introduces GFT, a method that combines imitation learning and reward fine-tuning with unbiased group advantages and dynamic coefficient rectification. The repository's popularity can be attributed to the growing interest in developing more efficient and effective methods for training large language models.
SUM-INNOVATION/RUMUS has seen moderate growth with a score of 9.55 and 190 stars, as it provides a Rust-based framework for training neural networks. While the repository's growth may not be as rapid as others, its popularity can be attributed to the increasing interest in using Rust for deep learning tasks.
WillowHe/EvoOpt_oppangu_optimization_model has maintained a steady growth with a score of 9.22 and 514 stars, as it provides solutions leveraging Openpangu-7B as the base model for fine-tuning and application of large language models in operations research optimization tasks. The repository's popularity can be attributed to its unique focus on applying large language models to specific domains.
semidark/kokoro-deutsch has seen modest growth with a score of 6.67 and 30 stars, as it provides a complete training recipe for fine-tuning Kokoro-82M on the German language. The repository's popularity can be attributed to its niche focus on language-specific fine-tuning tasks.
Dynamis-Labs/spectralquant has gained some attention with a score of 4.22 and 128 stars, as it introduces a method for breaking TurboQuant's compression limit via spectral structure. While the repository's growth may not be as rapid as others, its popularity can be attributed to the growing interest in exploring new methods for model compression.
Mintzs/oogaboogalm has seen limited growth with a score of 3.33 and 45 stars, but its unique approach to fine-tuning AI models by baking caveman system prompts into the model itself may still attract attention from researchers looking for innovative solutions.
PentesterFlow/OffensiveSET has maintained a steady presence with a score of 3.12 and 72 stars, as it provides an MCP server for generating high-quality pentesting conversation datasets for LLM fine-tuning. The repository's popularity can be attributed to its niche focus on security-related applications of large language models.
Other notable mentions include repositories that provide user-friendly tools or innovative methods for fine-tuning and training large language models, such as UNfukashigi/Anima-LoRA-Factory and ZJU-OmniAI/GFT.
mattmireles/gemma-tuner-multimodal has taken the top spot with a growth score of 54.55 and 1,391 stars, as it enables users to fine-tune Gemma 4 and 3n models on Apple Silicon using PyTorch and Metal Performance Shaders, making it an attractive option for those looking to leverage multimodal capabilities. The repository's high commit activity over the past month (100 commits) suggests a strong interest in this specific use case.
QingGo/engram-peft has gained significant attention with a growth score of 20.19 and 31 stars, as it provides an unofficial implementation of DeepSeek Engram, allowing users to inject high-capacity conditional memory into large language models via sparse retrieval PEFT without increasing inference FLOPs. The repository's popularity can be attributed to the growing interest in parameter-efficient methods for fine-tuning large language models.
UNfukashigi/Anima-LoRA-Factory has seen notable growth with a score of 15.78 and 23 stars, as it offers a user-friendly GUI tool designed specifically for training LoRAs for next-generation Anima diffusion models. The repository's popularity can be attributed to the increasing demand for specialized tools that simplify the process of fine-tuning large language models.
ZJU-OmniAI/GFT has gained traction with a growth score of 12.71 and 29 stars, as it introduces GFT, a method that combines imitation learning and reward fine-tuning with unbiased group advantages and dynamic coefficient rectification. The repository's popularity can be attributed to the growing interest in developing more efficient and effective methods for training large language models.
SUM-INNOVATION/RUMUS has seen moderate growth with a score of 9.55 and 190 stars, as it provides a Rust-based framework for training neural networks. While the repository's growth may not be as rapid as others, its popularity can be attributed to the increasing interest in using Rust for deep learning tasks.
WillowHe/EvoOpt_oppangu_optimization_model has maintained a steady growth with a score of 9.22 and 514 stars, as it provides solutions leveraging Openpangu-7B as the base model for fine-tuning and application of large language models in operations research optimization tasks. The repository's popularity can be attributed to its unique focus on applying large language models to specific domains.
semidark/kokoro-deutsch has seen modest growth with a score of 6.67 and 30 stars, as it provides a complete training recipe for fine-tuning Kokoro-82M on the German language. The repository's popularity can be attributed to its niche focus on language-specific fine-tuning tasks.
Dynamis-Labs/spectralquant has gained some attention with a score of 4.22 and 128 stars, as it introduces a method for breaking TurboQuant's compression limit via spectral structure. While the repository's growth may not be as rapid as others, its popularity can be attributed to the growing interest in exploring new methods for model compression.
Mintzs/oogaboogalm has seen limited growth with a score of 3.33 and 45 stars, but its unique approach to fine-tuning AI models by baking caveman system prompts into the model itself may still attract attention from researchers looking for innovative solutions.
PentesterFlow/OffensiveSET has maintained a steady presence with a score of 3.12 and 72 stars, as it provides an MCP server for generating high-quality pentesting conversation datasets for LLM fine-tuning. The repository's popularity can be attributed to its niche focus on security-related applications of large language models.
Other notable mentions include repositories that provide user-friendly tools or innovative methods for fine-tuning and training large language models, such as UNfukashigi/Anima-LoRA-Factory and ZJU-OmniAI/GFT.