README.md

We provide diverse examples about fine-tuning LLMs.

Make sure to execute these commands in the LLaMA-Factory directory.

Table of Contents

• LoRA Fine-Tuning
• QLoRA Fine-Tuning
• Full-Parameter Fine-Tuning
• Merging LoRA Adapters and Quantization
• Inferring LoRA Fine-Tuned Models
• Extras

Use CUDA_VISIBLE_DEVICES (GPU) or ASCEND_RT_VISIBLE_DEVICES (NPU) to choose computing devices.

Examples

LoRA Fine-Tuning

(Continuous) Pre-Training

llamafactory-cli train examples/train_lora/llama3_lora_pretrain.yaml

Supervised Fine-Tuning

llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml

Multimodal Supervised Fine-Tuning

llamafactory-cli train examples/train_lora/llava1_5_lora_sft.yaml
llamafactory-cli train examples/train_lora/qwen2vl_lora_sft.yaml

DPO/ORPO/SimPO Training

llamafactory-cli train examples/train_lora/llama3_lora_dpo.yaml

Multimodal DPO/ORPO/SimPO Training

llamafactory-cli train examples/train_lora/qwen2vl_lora_dpo.yaml

Reward Modeling

llamafactory-cli train examples/train_lora/llama3_lora_reward.yaml

PPO Training

llamafactory-cli train examples/train_lora/llama3_lora_ppo.yaml

KTO Training

llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml

Preprocess Dataset

It is useful for large dataset, use tokenized_path in config to load the preprocessed dataset.

llamafactory-cli train examples/train_lora/llama3_preprocess.yaml

Evaluating on MMLU/CMMLU/C-Eval Benchmarks

llamafactory-cli eval examples/train_lora/llama3_lora_eval.yaml

Batch Predicting and Computing BLEU and ROUGE Scores

llamafactory-cli train examples/train_lora/llama3_lora_predict.yaml

Supervised Fine-Tuning on Multiple Nodes

FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml

Supervised Fine-Tuning with DeepSpeed ZeRO-3 (Weight Sharding)

FORCE_TORCHRUN=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ds3.yaml

QLoRA Fine-Tuning

Supervised Fine-Tuning with 4/8-bit Bitsandbytes/HQQ/EETQ Quantization (Recommended)

llamafactory-cli train examples/train_qlora/llama3_lora_sft_otfq.yaml

Supervised Fine-Tuning with 4/8-bit GPTQ Quantization

llamafactory-cli train examples/train_qlora/llama3_lora_sft_gptq.yaml

Supervised Fine-Tuning with 4-bit AWQ Quantization

llamafactory-cli train examples/train_qlora/llama3_lora_sft_awq.yaml

Supervised Fine-Tuning with 2-bit AQLM Quantization

llamafactory-cli train examples/train_qlora/llama3_lora_sft_aqlm.yaml

Full-Parameter Fine-Tuning

Supervised Fine-Tuning on Single Node

FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml

Supervised Fine-Tuning on Multiple Nodes

FORCE_TORCHRUN=1 NNODES=2 RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml
FORCE_TORCHRUN=1 NNODES=2 RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml

Multimodal Supervised Fine-Tuning

FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen2vl_full_sft.yaml

Batch Predicting and Computing BLEU and ROUGE Scores

llamafactory-cli train examples/train_full/llama3_full_predict.yaml

Merging LoRA Adapters and Quantization

Merge LoRA Adapters

Note: DO NOT use quantized model or quantization_bit when merging LoRA adapters.

llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml

Quantizing Model using AutoGPTQ

llamafactory-cli export examples/merge_lora/llama3_gptq.yaml

Inferring LoRA Fine-Tuned Models

Use CLI

llamafactory-cli chat examples/inference/llama3_lora_sft.yaml

Use Web UI

llamafactory-cli webchat examples/inference/llama3_lora_sft.yaml

Launch OpenAI-style API

llamafactory-cli api examples/inference/llama3_lora_sft.yaml

Extras

Full-Parameter Fine-Tuning using GaLore

llamafactory-cli train examples/extras/galore/llama3_full_sft.yaml

Full-Parameter Fine-Tuning using BAdam

llamafactory-cli train examples/extras/badam/llama3_full_sft.yaml

Full-Parameter Fine-Tuning using Adam-mini

llamafactory-cli train examples/extras/adam_mini/qwen2_full_sft.yaml

LoRA+ Fine-Tuning

llamafactory-cli train examples/extras/loraplus/llama3_lora_sft.yaml

PiSSA Fine-Tuning

llamafactory-cli train examples/extras/pissa/llama3_lora_sft.yaml

Mixture-of-Depths Fine-Tuning

llamafactory-cli train examples/extras/mod/llama3_full_sft.yaml

LLaMA-Pro Fine-Tuning

bash examples/extras/llama_pro/expand.sh
llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml

FSDP+QLoRA Fine-Tuning

bash examples/extras/fsdp_qlora/train.sh