Today's Fine-tuning & Training: Fastest-Growing Projects — April 30, 2026
Today's the Fine-tuning & Training space, we're seeing a surge in innovative tools and repositories that are pushing the boundaries of language model optimization. Many of these projects focus on fine-tuning large language models (LLMs) for specific tasks or domains, while others explore new methods for efficient training and deployment. As AI continues to advance, these types of tools will play an increasingly important role in unlocking its full potential.
One of the fastest-growing repositories this week is mattmireles/gemma-tuner-multimodal, with a growth score of 49.98 and over 1,396 stars. This project allows 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 optimize their LLMs for multimodal tasks. Its rapid growth can be attributed to the increasing demand for efficient and specialized AI training tools.
Another notable repository is QingGo/engram-peft, which boasts a growth score of 17.94 and 31 stars. This unofficial implementation of DeepSeek Engram enables users to inject high-capacity conditional memory into LLMs via sparse retrieval PEFT without increasing inference FLOPs. Its growing popularity can be attributed to the interest in exploring new methods for efficient LLM training and deployment.
UNfukashigi/Anima-LoRA-Factory, with a growth score of 12.91 and 23 stars, is another project gaining traction this week. This user-friendly GUI tool allows users to train LoRAs for next-generation Anima diffusion models, making it an attractive option for those looking to explore new frontiers in AI-generated content. Its growing popularity can be attributed to the increasing interest in creative applications of AI.
ZJU-OmniAI/GFT, with a growth score of 10.89 and 29 stars, is another repository worth mentioning. This project introduces GFT, a method for fine-tuning LLMs using unbiased group advantages and dynamic coefficient rectification. Its growing popularity can be attributed to the interest in exploring new methods for efficient and effective LLM training.
We're also seeing growth in semidark/kokoro-deutsch, with a growth score of 6.17 and 31 stars. This project provides a complete training recipe for fine-tuning Kokoro-82M on German language tasks, making it an attractive option for those looking to optimize their LLMs for specific languages or domains. Its growing popularity can be attributed to the increasing demand for language-specific AI solutions.
Other notable repositories include Dynamis-Labs/spectralquant (growth score: 3.90, stars: 129), which explores new methods for efficient model compression; Mintzs/oogaboogalm (growth score: 3.00, stars: 45), which introduces a novel approach to reducing token use in LLMs; and PentesterFlow/OffensiveSET (growth score: 2.91, stars: 73), which provides a dataset generator for generating high-quality pentesting conversation datasets.
Finally, Goekdeniz-Guelmez/moshi-finetune-mlx (growth score: 0.69, stars: 25) is another project worth mentioning. This repository allows users to fine-tune Moshi models on Apple Silicon, making it an attractive option for those looking to optimize their speech-to-speech models. While its growth score may be lower than some of the other repositories mentioned, its unique focus on real-time speech-to-speech models sets it apart from other projects in the Fine-tuning & Training space.
One of the fastest-growing repositories this week is mattmireles/gemma-tuner-multimodal, with a growth score of 49.98 and over 1,396 stars. This project allows 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 optimize their LLMs for multimodal tasks. Its rapid growth can be attributed to the increasing demand for efficient and specialized AI training tools.
Another notable repository is QingGo/engram-peft, which boasts a growth score of 17.94 and 31 stars. This unofficial implementation of DeepSeek Engram enables users to inject high-capacity conditional memory into LLMs via sparse retrieval PEFT without increasing inference FLOPs. Its growing popularity can be attributed to the interest in exploring new methods for efficient LLM training and deployment.
UNfukashigi/Anima-LoRA-Factory, with a growth score of 12.91 and 23 stars, is another project gaining traction this week. This user-friendly GUI tool allows users to train LoRAs for next-generation Anima diffusion models, making it an attractive option for those looking to explore new frontiers in AI-generated content. Its growing popularity can be attributed to the increasing interest in creative applications of AI.
ZJU-OmniAI/GFT, with a growth score of 10.89 and 29 stars, is another repository worth mentioning. This project introduces GFT, a method for fine-tuning LLMs using unbiased group advantages and dynamic coefficient rectification. Its growing popularity can be attributed to the interest in exploring new methods for efficient and effective LLM training.
We're also seeing growth in semidark/kokoro-deutsch, with a growth score of 6.17 and 31 stars. This project provides a complete training recipe for fine-tuning Kokoro-82M on German language tasks, making it an attractive option for those looking to optimize their LLMs for specific languages or domains. Its growing popularity can be attributed to the increasing demand for language-specific AI solutions.
Other notable repositories include Dynamis-Labs/spectralquant (growth score: 3.90, stars: 129), which explores new methods for efficient model compression; Mintzs/oogaboogalm (growth score: 3.00, stars: 45), which introduces a novel approach to reducing token use in LLMs; and PentesterFlow/OffensiveSET (growth score: 2.91, stars: 73), which provides a dataset generator for generating high-quality pentesting conversation datasets.
Finally, Goekdeniz-Guelmez/moshi-finetune-mlx (growth score: 0.69, stars: 25) is another project worth mentioning. This repository allows users to fine-tune Moshi models on Apple Silicon, making it an attractive option for those looking to optimize their speech-to-speech models. While its growth score may be lower than some of the other repositories mentioned, its unique focus on real-time speech-to-speech models sets it apart from other projects in the Fine-tuning & Training space.