Welcome to the torchtune Documentation

torchtune is a Native-PyTorch library for LLM fine-tuning.

Getting Started

Topics in this section will help you get started with torchtune.

What is torchtune?

A gentle introduction to torchtune and how you can use the library in your projects.

overview.html

Installation instructions

A step-by-step tutorial on how to install torchtune.

install.html

Finetune your first model

Follow a simple tutorial to finetune Llama2 with torchtune.

tutorials/first_finetune_tutorial.html

Tutorials

Ready to experiment? Check out some of the interactive torchtune tutorials.

---

Llama3 in torchtune

Topics: finetuning,llama3

Finetuning with LoRA in torchtune

Topics: finetuning,llama2,lora

Parameter-efficient finetuning of Llama2 using LoRA

Understanding QLoRA in torchtune

Topics: finetuning,llama2,qlora

Using QLoRA to quantize base model weights and maximize memory savings

Finetuning with QAT in torchtune

Topics: finetuning,llama3,qat,quantization,evals

Finetuning of Llama3 using QAT

End-to-End Workflow with torchtune

Topics: finetuning,quantization,inference,evals,llama2

Train, Evaluate, Quantize and then Generate with your LLM.

Distilling Llama3 8B into 1B

Topics: finetuning,llama3,kd

Distilling Llama3 8B into 1B using Knowledge Distillation