TorchTune

Config based training (

Declarative Programming)

pip install torchtune

• FSDP

tune run --nproc_per_node 2 full_finetune_distributed --config llama2/7B_full

• LoRA,
  QLoRA

tune run lora_finetune_single_device \
--config llama2/7B_lora_single_device \
batch_size=8 \
enable_activation_checkpointing=True \
max_steps_per_epoch=128

• DPO

Config templates

torchtune/recipes/configs at main · pytorch/torchtune

A Native-PyTorch Library for LLM Fine-tuning. Contribute to pytorch/torchtune development by creating an account on GitHub.

https://github.com/pytorch/torchtune/tree/main/recipes/configs