Today's Fine-tuning & Training: Fastest-Growing Projects — May 06, 2026
This week, the Fine-tuning & Training space on GitHub has seen a surge in innovative tools and repositories focused on optimizing language models and improving their performance. Notably, many of these projects are leveraging techniques like parameter-efficient fine-tuning (PEFT) and sparse retrieval to inject high-capacity conditional memory into large language models (LLMs). As a result, we're seeing significant growth in repositories that offer user-friendly tools for training and fine-tuning LLMs.
mattmireles/gemma-tuner-multimodal is taking the lead with an impressive Growth Score of 39.88 and 1,407 stars. This repository provides a tool for fine-tuning Gemma 4 and 3n models with audio, images, and text on Apple Silicon using PyTorch and Metal Performance Shaders. Its rapid growth can be attributed to its unique approach to multimodal learning, which is gaining popularity in the AI research community.
QingGo/engram-peft has also seen significant traction with a Growth Score of 13.46 and 31 stars. This unofficial implementation of DeepSeek Engram injects high-capacity conditional memory into LLMs via sparse retrieval PEFT without increasing inference FLOPs. Its growth is likely due to the increasing interest in efficient fine-tuning methods that can improve LLM performance without sacrificing speed.
UNfukashigi/Anima-LoRA-Factory has gained popularity with a Growth Score of 8.82 and 27 stars, offering a user-friendly GUI tool for training LoRAs for next-generation Anima diffusion models. Its growth can be attributed to the growing interest in LoRA-based fine-tuning methods and the need for accessible tools that simplify this process.
ZJU-OmniAI/GFT has seen steady growth with a Growth Score of 7.65 and 30 stars, presenting a novel approach to fine-tuning LLMs using unbiased group advantages and dynamic coefficient rectification. Its growth is likely due to its innovative methodology, which addresses common challenges in reward fine-tuning.
raiyanyahya/how-to-train-your-gpt has maintained its popularity with a Growth Score of 5.09 and 436 stars, providing an extensively commented guide on building a modern LLM from scratch. Its enduring growth is likely due to the continued interest in understanding the fundamentals of LLM training and the need for accessible educational resources.
semidark/kokoro-deutsch has seen moderate growth with a Growth Score of 5.06 and 32 stars, offering a complete training recipe for fine-tuning Kokoro-82M on German language tasks. Its growth can be attributed to the increasing interest in multilingual LLMs and the need for high-quality training data.
Jackohhhh/MedLLM-Finetuning has gained attention with a Growth Score of 3.67 and 21 stars, providing an out-of-the-box fine-tuning framework for medical binary classification tasks. Its growth is likely due to the growing interest in applying LLMs to specialized domains like medicine.
Mintzs/oogaboogalm has seen slower but steady growth with a Growth Score of 2.35 and 47 stars, exploring the idea of baking caveman system prompts into LLMs through fine-tuning. Its growth is likely due to its novelty and the ongoing interest in improving LLM efficiency.
hlpun/Train-in-Silence has maintained a small but dedicated following with a Growth Score of 1.19 and 33 stars, offering a task-aware MCP server and automated VRAM calculator for LLM fine-tuning. Its growth is likely due to its unique value proposition, which simplifies the process of finding affordable GPUs for training.
mattmireles/gemma-tuner-multimodal is taking the lead with an impressive Growth Score of 39.88 and 1,407 stars. This repository provides a tool for fine-tuning Gemma 4 and 3n models with audio, images, and text on Apple Silicon using PyTorch and Metal Performance Shaders. Its rapid growth can be attributed to its unique approach to multimodal learning, which is gaining popularity in the AI research community.
QingGo/engram-peft has also seen significant traction with a Growth Score of 13.46 and 31 stars. This unofficial implementation of DeepSeek Engram injects high-capacity conditional memory into LLMs via sparse retrieval PEFT without increasing inference FLOPs. Its growth is likely due to the increasing interest in efficient fine-tuning methods that can improve LLM performance without sacrificing speed.
UNfukashigi/Anima-LoRA-Factory has gained popularity with a Growth Score of 8.82 and 27 stars, offering a user-friendly GUI tool for training LoRAs for next-generation Anima diffusion models. Its growth can be attributed to the growing interest in LoRA-based fine-tuning methods and the need for accessible tools that simplify this process.
ZJU-OmniAI/GFT has seen steady growth with a Growth Score of 7.65 and 30 stars, presenting a novel approach to fine-tuning LLMs using unbiased group advantages and dynamic coefficient rectification. Its growth is likely due to its innovative methodology, which addresses common challenges in reward fine-tuning.
raiyanyahya/how-to-train-your-gpt has maintained its popularity with a Growth Score of 5.09 and 436 stars, providing an extensively commented guide on building a modern LLM from scratch. Its enduring growth is likely due to the continued interest in understanding the fundamentals of LLM training and the need for accessible educational resources.
semidark/kokoro-deutsch has seen moderate growth with a Growth Score of 5.06 and 32 stars, offering a complete training recipe for fine-tuning Kokoro-82M on German language tasks. Its growth can be attributed to the increasing interest in multilingual LLMs and the need for high-quality training data.
Jackohhhh/MedLLM-Finetuning has gained attention with a Growth Score of 3.67 and 21 stars, providing an out-of-the-box fine-tuning framework for medical binary classification tasks. Its growth is likely due to the growing interest in applying LLMs to specialized domains like medicine.
Mintzs/oogaboogalm has seen slower but steady growth with a Growth Score of 2.35 and 47 stars, exploring the idea of baking caveman system prompts into LLMs through fine-tuning. Its growth is likely due to its novelty and the ongoing interest in improving LLM efficiency.
hlpun/Train-in-Silence has maintained a small but dedicated following with a Growth Score of 1.19 and 33 stars, offering a task-aware MCP server and automated VRAM calculator for LLM fine-tuning. Its growth is likely due to its unique value proposition, which simplifies the process of finding affordable GPUs for training.