aglio-lab/llm-fine-tuning-tools

This repository is a curated reference list rather than a piece of software to install. It catalogs tools used to fine-tune large language models, meaning taking a model that was already trained on general data and further training it on specific task or domain data so it performs better at a narrower job. The list holds 155 entries across 15 categories, covering both open source and commercial options, and it says it is reviewed monthly. The categories walk through the whole fine-tuning process. There are complete end-to-end frameworks that handle loading a model, formatting a dataset, running the training recipe, and exporting the result. There are libraries focused specifically on supervised fine-tuning, and on parameter-efficient methods like LoRA and QLoRA, which adjust a small number of extra parameters instead of retraining the whole model to save time and computing cost. Other sections cover scaling training across multiple GPUs, techniques like RLHF and DPO that train a model to follow human preferences, tools for cleaning and labeling training data, generating synthetic training examples, tracking experiment runs, searching for good training settings automatically, shrinking a trained model through quantization, converting and serving a finished checkpoint, and managed cloud services that handle fine-tuning for you. Every entry links to its original source, such as a GitHub repository or official documentation, so readers can check the claims for themselves. The list is also available as a machine-readable JSON file and CSV file for programmatic use. It documents its own methodology for how entries are selected and ordered, and keeps a section for tools that have been discontinued or replaced by newer projects rather than removing them from the record. The full README is longer than what was shown.