Enxin's picture
Upload folder using huggingface_hub
96fe658 verified

SWIFT安装

Wheel包安装

可以使用pip进行安装:

# 推荐
pip install 'ms-swift'
# 使用评测
pip install 'ms-swift[eval]' -U
# 全能力
pip install 'ms-swift[all]' -U

源代码安装

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

# 全能力
# pip install "git+https://github.com/modelscope/ms-swift.git#egg=ms-swift[all]"

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

# 全能力
# pip install -e '.[all]'

旧版本

SWIFT在3.0版本开始进行了不兼容式重构,如果需要使用2.x旧版本,请执行如下命令进行安装:

pip install ms-swift==2.*

镜像

# swift3.6.3
modelscope-registry.cn-hangzhou.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.28.1-swift3.6.3
modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.28.1-swift3.6.3
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.28.1-swift3.6.3

# swift3.5.3
modelscope-registry.cn-hangzhou.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.27.1-swift3.5.3
modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.27.1-swift3.5.3
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.27.1-swift3.5.3

# swift3.4.1.post1
modelscope-registry.cn-hangzhou.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.5.post1-modelscope1.26.0-swift3.4.1.post1
modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.5.post1-modelscope1.26.0-swift3.4.1.post1
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.5.post1-modelscope1.26.0-swift3.4.1.post1
历史镜像
# swift3.3.0.post1
modelscope-registry.cn-hangzhou.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.3-modelscope1.25.0-swift3.3.0.post1
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.3-modelscope1.25.0-swift3.3.0.post1

# swift3.2.2
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.5.1-modelscope1.25.0-swift3.2.2

更多镜像可以查看这里

支持的硬件

硬件环境 备注
A10/A100/H100
RTX20/30/40系列
T4/V100 部分模型出现NAN
Ascend NPU 部分模型出现NAN或算子不支持
MPS 参考issue 4572
CPU

运行环境

范围 推荐 备注
python >=3.9 3.10
cuda cuda12 使用cpu、npu、mps则无需安装
torch >=2.0
transformers >=4.33 4.51.3
modelscope >=1.23
peft >=0.11,<0.16
trl >=0.13,<0.19 0.18 RLHF
deepspeed >=0.14 0.16.9 训练
vllm >=0.5.1 0.8.5.post1 推理/部署/评测
sglang 0.4.6.post5 推理/部署/评测
lmdeploy >=0.5,<0.9 0.8 推理/部署/评测
evalscope >=0.11 评测

更多可选依赖可以参考这里

Notebook环境

Swift支持训练的绝大多数模型都可以在A10显卡上使用,用户可以使用ModelScope官方提供的免费显卡资源:

  1. 进入ModelScope官方网站并登录。
  2. 点击左侧的我的Notebook并开启一个免费GPU实例。
  3. 愉快地薅A10显卡羊毛。