torchtune

A playful project that cleverly transforms everyday sentences into their mischievous "toxic" counterparts. 😈

Finetune LLM using Torchtune on Windows

Internet-scale data-parallel fine-tuning of Qwen2-0.5B-Instruct using Hivemind + TorchTune. Initial peer on public IP; second peers on free GPUs (e.g., Kaggle).

This repository contains a comprehensive pipeline for fine-tuning Large Language Models (LLMs) to optimize their performance for specialized healthcare applications. The project focuses on transforming a base Llama model into a specialized medical assistant capable of classifying patient intent and answering clinical queries.

Release engineering desk for LLM fine-tuning, evaluation drift, and rollout readiness.

PyTorch-native fine-tuning recipes — LoRA, QLoRA, knowledge distillation, DPO

Local-first LLM fine-tuning workstation with a modern UI — supports Unsloth, Axolotl, and Torchtune backends with real-time GPU telemetry, live loss curves, and VRAM pre-flight checks.

This repository provides resources for fine-tuning various types of models using different techniques and frameworks.