PullRepo

Daily radar for the fastest-growing AI tools & repos

Today's Fine-tuning & Training: Fastest-Growing Projects — November 22, 2026

Today's Fine-tuning & Training category highlights a diverse range of projects that cater to various use cases, from fine-tuning language models for specific tasks like PCB layout planning and Cantonese speech recognition to training agents on personalized data sets. The standout project is "decision-forge," which continues its impressive growth trajectory by compiling organizational policies into verifiable decision-making datasets.

oguzeldereli/ttf-home
This repository allows users to run, train, convert, fine-tune, and export ternary language models directly from their home computers using GGUF. Its Growth Score of 24.69 reflects its rapid adoption among developers interested in low-resource machine learning setups for language modeling.

Juneo-Lab/decision-forge
Decision-forge compiles organizational policy into verifiable decision training data, enabling users to create datasets that can be used to train AI models to make decisions based on company policies and regulations. With a Growth Score of 22.39 and over 350 stars, the project's popularity underscores its relevance in regulatory compliance and ethical AI development.

drost-ai/tinker-ai-skill
This skill allows users to fine-tune an agentic model on their own Claude Code sessions within environments like Claude Code, Codex, and OpenCode. The repository has garnered 330 stars, indicating a strong community interest in enhancing the capabilities of AI agents through personalized training data.

Ssbullock/pcb-layout-llm
Fine-tuned for KiCad PCB architectural layout planning and routing strategy, this Qwen2.5-Coder-32B model is designed to assist engineers in creating efficient board layouts. Despite fewer recent commits, the repository's Growth Score of 10.90 suggests steady interest from hardware design professionals looking to leverage AI in their workflows.

koelony/qwen3-asr-cantonese-hk
This project focuses on fine-tuning Qwen3-ASR-1.7B for Cantonese speech recognition, providing GGUF files and benchmarks suitable for use with llama.cpp or OraType. With 77 stars, the repository shows promise in addressing language-specific challenges in natural language processing.

Zihao-Wu06/distillops
Distillops is a tool designed to distill black-box LLMs into more accessible models using rejection sampling and fine-tuning techniques. The project's Growth Score of 6.80, coupled with over 50 recent commits in the past month, indicates active development and growing interest from researchers and practitioners.

ninetyhe-90/training-explore
This interactive explorer offers a hands-on approach to learning how large language models are trained, making it an educational resource for those new to AI model training. Although only one commit was made recently, the repository's 155 stars suggest it has found a dedicated user base interested in understanding and experimenting with LLM training methods.

sous-labs/songe
Songe is a local model training engine designed to utilize idle compute resources effectively for model training tasks. With over 160 stars and frequent commits, the project demonstrates ongoing development and growing interest from users looking to optimize their computational resources efficiently.

Jaxsonb04/so101-pi05
This repository focuses on building SO-101 robot arms and fine-tuning π0.5 models on these robots using LeRobot. Despite fewer recent commits, the 28 stars indicate a niche but engaged community interested in robotics and AI integration.

EDAPINENUT/fewsteprl-sd35
The project aims to fine-tune few-step diffusion models with reinforcement learning techniques for SD-3.5. With 57 stars and steady contributions, the repository is gaining traction among researchers exploring advanced model training strategies in generative modeling.

These projects showcase a range of innovative approaches to Fine-tuning & Training across different domains, from language processing to robotics, highlighting the dynamic nature of AI development today.
Back to all reports