GitHub - modelscope/ms-swift: Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 500+ LLMs (Qwen3, Qwen3-MoE, Llama4, InternLM3, DeepSeek-R1, ...) and 200+ MLLMs (Qwen2.5-VL, Qwen2.5-Omni, Qwen2-Audio, Ovis2, InternVL3, Llava, GLM4v, Phi4, ...) (AAAI 2025). (original) (raw)

SWIFT (Scalable lightWeight Infrastructure for Fine-Tuning)

ModelScope Community Website
中文 ｜ English

Paper ｜ English Documentation ｜ 中文文档

📖 Table of Contents

• Groups
• Introduction
• News
• Installation
• Quick Start
• Usage
• License
• Citation

☎ Groups

You can contact us and communicate with us by adding our group:

Discord Group	WeChat Group

📝 Introduction

🍲 ms-swift is an official framework provided by the ModelScope community for fine-tuning and deploying large language models and multi-modal large models. It currently supports the training (pre-training, fine-tuning, human alignment), inference, evaluation, quantization, and deployment of 500+ large models and 200+ multi-modal large models. These large language models (LLMs) include models such as Qwen3, Qwen3-MoE, Qwen2.5, InternLM3, GLM4, Mistral, DeepSeek-R1, Yi1.5, TeleChat2, Baichuan2, and Gemma2. The multi-modal LLMs include models such as Qwen2.5-VL, Qwen2-Audio, Llama4, Llava, InternVL3, MiniCPM-V-2.6, GLM4v, Xcomposer2.5, Yi-VL, DeepSeek-VL2, Phi3.5-Vision, and GOT-OCR2.

🍔 Additionally, ms-swift incorporates the latest training technologies, including lightweight techniques such as LoRA, QLoRA, Llama-Pro, LongLoRA, GaLore, Q-GaLore, LoRA+, LISA, DoRA, FourierFt, ReFT, UnSloth, and Liger, as well as human alignment training methods like DPO, GRPO, RM, PPO, KTO, CPO, SimPO, and ORPO. ms-swift supports acceleration of inference, evaluation, and deployment modules using vLLM and LMDeploy, and it supports model quantization with technologies like GPTQ, AWQ, and BNB. Furthermore, ms-swift offers a Gradio-based Web UI and a wealth of best practices.

Why choose ms-swift?

• 🍎 Model Types: Supports 500+ pure text large models, 200+ multi-modal large models, as well as All-to-All multi-modal models, sequence classification models, and embedding models, covering the entire process from training to deployment.
• Dataset Types: Comes with 150+ pre-training, fine-tuning, human alignment, multi-modal datasets, and supports custom datasets.
• Hardware Support: Compatible with CPU, RTX series, T4/V100, A10/A100/H100, Ascend NPU, MPS, etc.
• 🍊 Lightweight Training: Supports lightweight fine-tuning methods like LoRA, QLoRA, DoRA, LoRA+, ReFT, RS-LoRA, LLaMAPro, Adapter, GaLore, Q-Galore, LISA, UnSloth, Liger-Kernel.
• Distributed Training: Supports distributed data parallel (DDP), device_map simple model parallelism, DeepSpeed ZeRO2/ZeRO3, FSDP, and other distributed training techniques.
• Quantization Training: Supports training quantized models like BNB, AWQ, GPTQ, AQLM, HQQ, EETQ.
• RLHF Training: Supports human alignment training methods such as DPO, GRPO, RM, PPO, KTO, CPO, SimPO, ORPO for both pure text and multi-modal large models.
• 🍓 Multi-Modal Training: Supports training on different modalities like images, videos, and audio, for tasks like VQA, captioning, OCR, and grounding.
• Interface Training: Provides capabilities for training, inference, evaluation, quantization through an interface, completing the whole large model pipeline.
• Plugin and Extension: Supports custom model and dataset extensions, as well as customization of components like loss, metric, trainer, loss-scale, callback, optimizer.
• 🍉 Toolbox Capabilities: Offers not only training support for large models and multi-modal large models but also covers the entire process of inference, evaluation, quantization, and deployment.
• Inference Acceleration: Supports inference acceleration engines like PyTorch, vLLM, LmDeploy, and provides OpenAI API for accelerating inference, deployment, and evaluation modules.
• Model Evaluation: Uses EvalScope as the evaluation backend and supports evaluation on 100+ datasets for both pure text and multi-modal models.
• Model Quantization: Supports AWQ, GPTQ, and BNB quantized exports, with models that can use vLLM/LmDeploy for inference acceleration and continue training.

🎉 News

• 🎁 2025.06.11: Support for using Megatron parallelism techniques for RLHF training. The training script can be found here.
• 🎁 2025.05.29: Support sequence parallel in pt, sft, dpo and grpo, check script here.
• 🎁 2025.05.11: GRPO now supports custom processing logic for reward models. See the GenRM example here.
• 🎁 2025.04.15: The ms-swift paper has been accepted by AAAI 2025. You can find the paper at this link.
• 🎁 2025.03.23: Multi-round GRPO is now supported for training multi-turn dialogue scenarios (e.g., agent tool calling). Please refer to the training script.
• 🎁 2025.03.16: Support for Megatron's parallel training techniques is now available. Please see the Megatron-SWIFT training documentation.
• 🎁 2025.03.15: Fine-tuning of embedding models for both pure text and multimodal models is supported. Please check the training script.
• 🎁 2025.03.05: The hybrid mode for GRPO is supported, with a script for training a 72B model on 4 GPUs (4*80G) available here. Tensor parallelism with vllm is also supported, with the training script available here.
• 🎁 2025.02.21: The GRPO algorithm now supports LMDeploy, with the training script available here. Additionally, the performance of the GRPO algorithm has been tested, achieving a training speed increase of up to 300% using various tricks. Please check the WanDB table here.
• 🎁 2025.02.21: The swift sample command is now supported. The reinforcement fine-tuning script can be found here, and the large model API distillation sampling script is available here.
• 🔥 2025.02.12: Support for the GRPO (Group Relative Policy Optimization) training algorithm has been added. Documentation is available here.
• 🎁 2024.12.04: Major update to ms-swift 3.0. Please refer to the release notes and changes. More
• 🎉 2024.08.12: The ms-swift paper has been published on arXiv and can be read here.
• 🔥 2024.08.05: Support for using evalscope as a backend for evaluating large models and multimodal models.
• 🔥 2024.07.29: Support for using vllm and lmdeploy to accelerate inference for large models and multimodal models. When performing infer/deploy/eval, you can specify --infer_backend vllm/lmdeploy.
• 🔥 2024.07.24: Support for human preference alignment training for multimodal large models, including DPO/ORPO/SimPO/CPO/KTO/RM/PPO.
• 🔥 2024.02.01: Support for Agent training! The training algorithm is derived from this paper.

🛠️ Installation

To install using pip:

To install from source:

pip install git+https://github.com/modelscope/ms-swift.git

git clone https://github.com/modelscope/ms-swift.git cd ms-swift pip install -e .

Running Environment:

For more optional dependencies, you can refer to here.

🚀 Quick Start

10 minutes of self-cognition fine-tuning of Qwen2.5-7B-Instruct on a single 3090 GPU:

Command Line Interface

22GB

CUDA_VISIBLE_DEVICES=0
swift sft
--model Qwen/Qwen2.5-7B-Instruct
--train_type lora
--dataset 'AI-ModelScope/alpaca-gpt4-data-zh#500'
'AI-ModelScope/alpaca-gpt4-data-en#500'
'swift/self-cognition#500'
--torch_dtype bfloat16
--num_train_epochs 1
--per_device_train_batch_size 1
--per_device_eval_batch_size 1
--learning_rate 1e-4
--lora_rank 8
--lora_alpha 32
--target_modules all-linear
--gradient_accumulation_steps 16
--eval_steps 50
--save_steps 50
--save_total_limit 2
--logging_steps 5
--max_length 2048
--output_dir output
--system 'You are a helpful assistant.'
--warmup_ratio 0.05
--dataloader_num_workers 4
--model_author swift
--model_name swift-robot

Tips:

• If you want to train with a custom dataset, you can refer to this guide to organize your dataset format and specify --dataset <dataset_path>.
• The --model_author and --model_name parameters are only effective when the dataset includes swift/self-cognition.
• To train with a different model, simply modify --model <model_id/model_path>.
• By default, ModelScope is used for downloading models and datasets. If you want to use HuggingFace, simply specify --use_hf true.

After training is complete, use the following command to infer with the trained weights:

• Here, --adapters should be replaced with the last checkpoint folder generated during training. Since the adapters folder contains the training parameter file args.json, there is no need to specify --model, --system separately; Swift will automatically read these parameters. To disable this behavior, you can set --load_args false.

Using an interactive command line for inference.

CUDA_VISIBLE_DEVICES=0
swift infer
--adapters output/vx-xxx/checkpoint-xxx
--stream true
--temperature 0
--max_new_tokens 2048

merge-lora and use vLLM for inference acceleration

CUDA_VISIBLE_DEVICES=0
swift infer
--adapters output/vx-xxx/checkpoint-xxx
--stream true
--merge_lora true
--infer_backend vllm
--max_model_len 8192
--temperature 0
--max_new_tokens 2048

Finally, use the following command to push the model to ModelScope:

CUDA_VISIBLE_DEVICES=0
swift export
--adapters output/vx-xxx/checkpoint-xxx
--push_to_hub true
--hub_model_id ''
--hub_token ''
--use_hf false

Web-UI

The Web-UI is a zero-threshold training and deployment interface solution based on Gradio interface technology. For more details, you can check here.

SWIFT_UI_LANG=en swift web-ui

Using Python

ms-swift also supports training and inference using Python. Below is pseudocode for training and inference. For more details, you can refer to here.

Training:

Retrieve the model and template, and add a trainable LoRA module

model, tokenizer = get_model_tokenizer(model_id_or_path, ...) template = get_template(model.model_meta.template, tokenizer, ...) model = Swift.prepare_model(model, lora_config)

Download and load the dataset, and encode the text into tokens

train_dataset, val_dataset = load_dataset(dataset_id_or_path, ...) train_dataset = EncodePreprocessor(template=template)(train_dataset, num_proc=num_proc) val_dataset = EncodePreprocessor(template=template)(val_dataset, num_proc=num_proc)

Train the model

trainer = Seq2SeqTrainer( model=model, args=training_args, data_collator=template.data_collator, train_dataset=train_dataset, eval_dataset=val_dataset, template=template, ) trainer.train()

Inference:

Perform inference using the native PyTorch engine

engine = PtEngine(model_id_or_path, adapters=[lora_checkpoint]) infer_request = InferRequest(messages=[{'role': 'user', 'content': 'who are you?'}]) request_config = RequestConfig(max_tokens=max_new_tokens, temperature=temperature)

resp_list = engine.infer([infer_request], request_config) print(f'response: {resp_list[0].choices[0].message.content}')

✨ Usage

Here is a minimal example of training to deployment using ms-swift. For more details, you can check the examples.

• If you want to use other models or datasets (including multimodal models and datasets), you only need to modify --model to specify the corresponding model's ID or path, and modify --dataset to specify the corresponding dataset's ID or path.
• By default, ModelScope is used for downloading models and datasets. If you want to use HuggingFace, simply specify --use_hf true.

Training

Supported Training Methods:

Pre-training:

8*A100

NPROC_PER_NODE=8
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
swift pt
--model Qwen/Qwen2.5-7B
--dataset swift/chinese-c4
--streaming true
--train_type full
--deepspeed zero2
--output_dir output
--max_steps 10000
...

Fine-tuning:

CUDA_VISIBLE_DEVICES=0 swift sft
--model Qwen/Qwen2.5-7B-Instruct
--dataset AI-ModelScope/alpaca-gpt4-data-en
--train_type lora
--output_dir output
...

RLHF:

CUDA_VISIBLE_DEVICES=0 swift rlhf
--rlhf_type dpo
--model Qwen/Qwen2.5-7B-Instruct
--dataset hjh0119/shareAI-Llama3-DPO-zh-en-emoji
--train_type lora
--output_dir output
...

Inference

CUDA_VISIBLE_DEVICES=0 swift infer
--model Qwen/Qwen2.5-7B-Instruct
--stream true
--infer_backend pt
--max_new_tokens 2048

LoRA

CUDA_VISIBLE_DEVICES=0 swift infer
--model Qwen/Qwen2.5-7B-Instruct
--adapters swift/test_lora
--stream true
--infer_backend pt
--temperature 0
--max_new_tokens 2048

Interface Inference

CUDA_VISIBLE_DEVICES=0 swift app
--model Qwen/Qwen2.5-7B-Instruct
--stream true
--infer_backend pt
--max_new_tokens 2048

Deployment

CUDA_VISIBLE_DEVICES=0 swift deploy
--model Qwen/Qwen2.5-7B-Instruct
--infer_backend vllm

Sampling

CUDA_VISIBLE_DEVICES=0 swift sample
--model LLM-Research/Meta-Llama-3.1-8B-Instruct
--sampler_engine pt
--num_return_sequences 5
--dataset AI-ModelScope/alpaca-gpt4-data-zh#5

Evaluation

CUDA_VISIBLE_DEVICES=0 swift eval
--model Qwen/Qwen2.5-7B-Instruct
--infer_backend lmdeploy
--eval_backend OpenCompass
--eval_dataset ARC_c

Quantization

CUDA_VISIBLE_DEVICES=0 swift export
--model Qwen/Qwen2.5-7B-Instruct
--quant_bits 4 --quant_method awq
--dataset AI-ModelScope/alpaca-gpt4-data-zh
--output_dir Qwen2.5-7B-Instruct-AWQ

Push Model

swift export
--model
--push_to_hub true
--hub_model_id ''
--hub_token ''

🏛 License

This framework is licensed under the Apache License (Version 2.0). For models and datasets, please refer to the original resource page and follow the corresponding License.

📎 Citation

@misc{zhao2024swiftascalablelightweightinfrastructure, title={SWIFT:A Scalable lightWeight Infrastructure for Fine-Tuning}, author={Yuze Zhao and Jintao Huang and Jinghan Hu and Xingjun Wang and Yunlin Mao and Daoze Zhang and Zeyinzi Jiang and Zhikai Wu and Baole Ai and Ang Wang and Wenmeng Zhou and Yingda Chen}, year={2024}, eprint={2408.05517}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2408.05517}, }

Star History