Today's Fine-tuning & Training: Fastest-Growing Projects — May 02, 2026
Today's the Fine-tuning & Training space, we're seeing a surge of innovative tools that cater to specific needs in AI model training. From fine-tuning multimodal models on Apple Silicon to injecting high-capacity conditional memory into large language models (LLMs), developers are pushing the boundaries of what's possible with AI training.
mattmireles/gemma-tuner-multimodal has taken the top spot this week, boasting a growth score of 46.12 and 1,400 stars. This tool allows users to fine-tune Gemma 4 and 3n models with audio, images, and text on Apple Silicon using PyTorch and Metal Performance Shaders, making it an attractive solution for those looking to optimize their multimodal model training.
QingGo/engram-peft has also seen significant growth, with a score of 16.15 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, making it an exciting development for those working with large language models.
UNfukashigi/Anima-LoRA-Factory has garnered a growth score of 11.42 and 24 stars, thanks to its user-friendly GUI tool designed specifically for training LoRAs for next-generation Anima diffusion models. This tool's popularity can be attributed to the growing interest in diffusion models and the need for more accessible training tools.
ZJU-OmniAI/GFT has secured a growth score of 9.53 and 29 stars, with its innovative approach to fine-tuning using unbiased group advantages and dynamic coefficient rectification. This tool's unique methodology is likely attracting attention from researchers and developers looking for new ways to optimize their training processes.
semidark/kokoro-deutsch has seen a modest growth score of 5.68 and 31 stars, but its documented training recipe for fine-tuning Kokoro-82M on the German language makes it an invaluable resource for those working with this specific model.
Dynamis-Labs/spectralquant boasts an impressive 131 stars, despite a relatively low growth score of 3.65. This tool's innovative approach to breaking TurboQuant's compression limit via spectral structure is likely attracting attention from researchers and developers in the field.
Mintzs/oogaboogalm has garnered a growth score of 2.75 and 46 stars, with its unique approach to fine-tuning AI models using caveman system prompts and skills to reduce token use. This tool's popularity can be attributed to the growing interest in optimizing language model performance.
hlpun/Train-in-Silence has seen limited growth this week, but its task-aware MCP server and automated VRAM calculator for LLM fine-tuning make it a valuable resource for those looking to optimize their training processes.
Finally, Goekdeniz-Guelmez/moshi-finetune-mlx rounds out the list with a growth score of 0.64 and 25 stars. This tool allows users to fine-tune Moshi models on Apple Silicon, catering to the growing demand for optimized speech-to-speech models.
mattmireles/gemma-tuner-multimodal has taken the top spot this week, boasting a growth score of 46.12 and 1,400 stars. This tool allows users to fine-tune Gemma 4 and 3n models with audio, images, and text on Apple Silicon using PyTorch and Metal Performance Shaders, making it an attractive solution for those looking to optimize their multimodal model training.
QingGo/engram-peft has also seen significant growth, with a score of 16.15 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, making it an exciting development for those working with large language models.
UNfukashigi/Anima-LoRA-Factory has garnered a growth score of 11.42 and 24 stars, thanks to its user-friendly GUI tool designed specifically for training LoRAs for next-generation Anima diffusion models. This tool's popularity can be attributed to the growing interest in diffusion models and the need for more accessible training tools.
ZJU-OmniAI/GFT has secured a growth score of 9.53 and 29 stars, with its innovative approach to fine-tuning using unbiased group advantages and dynamic coefficient rectification. This tool's unique methodology is likely attracting attention from researchers and developers looking for new ways to optimize their training processes.
semidark/kokoro-deutsch has seen a modest growth score of 5.68 and 31 stars, but its documented training recipe for fine-tuning Kokoro-82M on the German language makes it an invaluable resource for those working with this specific model.
Dynamis-Labs/spectralquant boasts an impressive 131 stars, despite a relatively low growth score of 3.65. This tool's innovative approach to breaking TurboQuant's compression limit via spectral structure is likely attracting attention from researchers and developers in the field.
Mintzs/oogaboogalm has garnered a growth score of 2.75 and 46 stars, with its unique approach to fine-tuning AI models using caveman system prompts and skills to reduce token use. This tool's popularity can be attributed to the growing interest in optimizing language model performance.
hlpun/Train-in-Silence has seen limited growth this week, but its task-aware MCP server and automated VRAM calculator for LLM fine-tuning make it a valuable resource for those looking to optimize their training processes.
Finally, Goekdeniz-Guelmez/moshi-finetune-mlx rounds out the list with a growth score of 0.64 and 25 stars. This tool allows users to fine-tune Moshi models on Apple Silicon, catering to the growing demand for optimized speech-to-speech models.