Today's Fine-tuning & Training: Fastest-Growing Projects — May 07, 2026
Today's the Fine-tuning & Training space, we're seeing a surge of interest in innovative approaches to improving language model performance. From unofficial implementations of cutting-edge architectures to user-friendly GUI tools for training LoRAs, developers are exploring new ways to fine-tune and train their models. With growth scores ranging from 0.91 to 12.92, these projects are gaining traction on GitHub.
QingGo's engram-peft repository takes the top spot with a staggering growth score of 12.92 and 31 stars. This unofficial implementation of DeepSeek Engram injects high-capacity conditional memory into LLMs via sparse retrieval PEFT without increasing inference FLOPs, making it an attractive solution for developers looking to boost their model's performance. Its rapid growth is likely due to the innovative approach it takes to fine-tuning.
UNfukashigi's Anima-LoRA-Factory comes in second with a growth score of 8.33 and 27 stars. This user-friendly GUI tool is designed specifically for training LoRAs for next-generation Anima diffusion models, making it an essential resource for developers working with these models. Its ease of use and specialized functionality are likely driving its popularity.
ZJU-OmniAI's GFT repository boasts a growth score of 7.29 and 30 stars. This project introduces a novel approach to fine-tuning, using unbiased group advantages and dynamic coefficient rectification to improve model performance. Its unique methodology is likely attracting attention from developers looking for new ways to optimize their models.
raiyanyahya's how-to-train-your-gpt repository may have a lower growth score of 5.17, but its impressive 624 stars demonstrate its enduring popularity. This comprehensive guide walks developers through building a modern LLM from scratch, with every line commented and explained in an approachable manner. Its continued growth is likely due to its value as a learning resource.
semidark's kokoro-deutsch repository has a growth score of 4.89 and 33 stars. This project provides a complete training recipe for fine-tuning Kokoro-82M on German, making it a valuable resource for developers working with this language. Its growth is likely driven by the increasing demand for multilingual models.
Jackohhhh's MedLLM-Finetuning repository boasts a growth score of 3.50 and 21 stars. This out-of-the-box LLM fine-tuning framework is specifically designed for medical binary classification tasks, making it an attractive solution for developers working in this domain. Its ease of use and specialized functionality are likely driving its popularity.
Mintzs' oogaboogalm repository has a growth score of 2.26 and 47 stars. This project explores the idea of fine-tuning models to reduce token usage, using caveman system prompts and skills as inspiration. Its unique approach is likely attracting attention from developers looking for innovative solutions.
hlpun's Train-in-Silence repository boasts a growth score of 1.20 and 36 stars. This Task-Aware MCP server and automated VRAM calculator helps developers instantly find the cheapest, fastest GPUs across multiple cloud providers for LLM fine-tuning. Its practical value is likely driving its growth.
vvvvvjdy's D-OPSD repository rounds out our list with a growth score of 0.91 and 21 stars. This official repo introduces On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models, offering a novel approach to fine-tuning. Its growth is likely driven by interest in this new methodology.
QingGo's engram-peft repository takes the top spot with a staggering growth score of 12.92 and 31 stars. This unofficial implementation of DeepSeek Engram injects high-capacity conditional memory into LLMs via sparse retrieval PEFT without increasing inference FLOPs, making it an attractive solution for developers looking to boost their model's performance. Its rapid growth is likely due to the innovative approach it takes to fine-tuning.
UNfukashigi's Anima-LoRA-Factory comes in second with a growth score of 8.33 and 27 stars. This user-friendly GUI tool is designed specifically for training LoRAs for next-generation Anima diffusion models, making it an essential resource for developers working with these models. Its ease of use and specialized functionality are likely driving its popularity.
ZJU-OmniAI's GFT repository boasts a growth score of 7.29 and 30 stars. This project introduces a novel approach to fine-tuning, using unbiased group advantages and dynamic coefficient rectification to improve model performance. Its unique methodology is likely attracting attention from developers looking for new ways to optimize their models.
raiyanyahya's how-to-train-your-gpt repository may have a lower growth score of 5.17, but its impressive 624 stars demonstrate its enduring popularity. This comprehensive guide walks developers through building a modern LLM from scratch, with every line commented and explained in an approachable manner. Its continued growth is likely due to its value as a learning resource.
semidark's kokoro-deutsch repository has a growth score of 4.89 and 33 stars. This project provides a complete training recipe for fine-tuning Kokoro-82M on German, making it a valuable resource for developers working with this language. Its growth is likely driven by the increasing demand for multilingual models.
Jackohhhh's MedLLM-Finetuning repository boasts a growth score of 3.50 and 21 stars. This out-of-the-box LLM fine-tuning framework is specifically designed for medical binary classification tasks, making it an attractive solution for developers working in this domain. Its ease of use and specialized functionality are likely driving its popularity.
Mintzs' oogaboogalm repository has a growth score of 2.26 and 47 stars. This project explores the idea of fine-tuning models to reduce token usage, using caveman system prompts and skills as inspiration. Its unique approach is likely attracting attention from developers looking for innovative solutions.
hlpun's Train-in-Silence repository boasts a growth score of 1.20 and 36 stars. This Task-Aware MCP server and automated VRAM calculator helps developers instantly find the cheapest, fastest GPUs across multiple cloud providers for LLM fine-tuning. Its practical value is likely driving its growth.
vvvvvjdy's D-OPSD repository rounds out our list with a growth score of 0.91 and 21 stars. This official repo introduces On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models, offering a novel approach to fine-tuning. Its growth is likely driven by interest in this new methodology.