Zzhangyihuiben【master】【feature】add Dockerfile

MindSpore

English | 中文

Quick Reference

MindSpore Transformers is maintained by the MindSpore Transformers Community 维护
Where to get help

MindSpore Transformers

The MindSpore Transformers suite aims to build a comprehensive development toolkit covering the entire lifecycle of large-scale models, including pre-training, fine-tuning, evaluation, inference, and deployment. It provides industry-leading large Transformer-based language models, multimodal understanding models, and omni-modal models, enabling users to easily achieve end-to-end development of large-scale models.

Supported Tags and Dockerfile Usage

Tag Specification

Tags follow this format:

 <MindSpore Transformers Version>-<Hardware Info (Chip)>-<Operating System>-<Python Version>

Field	Example Values	Description
MindSpore Transformers Version	1.8.0	Corresponds to the version identifier in MindSpore Transformers official release tags
Hardware Info (Chip)	910b / a3	Ascend chip model identifier
Operating System	ubuntu22.04 / openeuler24.03	Operating system distribution and version used in the base image
Python Version	py3.11	Major Python version built into the image

Tips: System architecture is automatically detected via Docker Manifest, no need to specify in the tag.

Image Repository Address

MindSpore Transformers Ascend images are hosted on Huawei Cloud SWR image repository:

swr.cn-south-1.myhuaweicloud.com/ascendhub/mindformers

Full Image Example：

swr.cn-south-1.myhuaweicloud.com/ascendhub/mindformers:1.8.0-910b-ubuntu22.04-py3.11

构建参数

Parameter	Description	Required	Source	Example Values
CANN_VERSION	Ascend CANN toolkit version	Yes	CANN image tag	8.5.0
CHIP_ARCH	Ascend chip architecture identifier	Yes	Tag specification	910b / a3
OS_SYSTEM	Base image operating system and version	Yes	Tag specification	ubuntu22.04 / openeuler24.03
PY_VERSION	Python version built into the base image	Yes	Tag specification	py3.11
MINDSPORE_VERSION	MindSpore version number	Yes	MindSpore repository releases	2.7.2
MINDSPORE_TRANSFORMERS_VERSION	MindSpore version number	Yes	MindSpore repository releases	1.8.0
PIP_INDEX_URL	pip installation source URL (default: Huawei Cloud mirror)	No	PyPI mirror source	https://mirrors.huaweicloud.com/repository/pypi/simple

Quick Start

Building MindSpore Image

docker build \
--build-arg CANN_VERSION=8.5.0 \
--build-arg CHIP_ARCH=910b \
--build-arg OS_SYSTEM=ubuntu22.04 \
--build-arg PY_VERSION=py3.11 \
--build-arg MINDSPORE_VERSION=2.7.2 \
--build-arg MINDFORMERS_VERSION=1.8.0 \
--build-arg PIP_INDEX_URL=https://mirrors.huaweicloud.com/repository/pypi/simple \
-t mindformers:1.8.0-910b-ubuntu22.04-py3.11 \
-f Dockerfile .

运行 MindSpore Transformers 容器

docker run \
    --privileged \
    --name mindformers_container \
    --device /dev/davinci1 \
    --device /dev/davinci_manager \
    --device /dev/devmm_svm \
    --device /dev/hisi_hdc \
    -v /usr/local/dcmi:/usr/local/dcmi \
    -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
    -v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
    -v /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info \
    -v /etc/ascend_install.info:/etc/ascend_install.info \
    -it mindspore:tag bash

Security Risks

When running MindSpore Transformers in Docker containers, pay attention to the following security risks:

Running as the root user:
Containers run as the root user by default, which may introduce security risks. It is recommended to create and use a non-privileged user in production environments.
Running with --privileged mode:
To ensure Ascend NPU functionality works properly, the container may require the --privileged option. This mode grants the container elevated access to host devices and system resources, increasing potential security risks. It is recommended to use it only in trusted environments and reduce risks through device whitelisting, resource limits, and network isolation.
Missing CPU and memory resource limits:
Without resource limits, containers may consume excessive system resources and impact host performance. It is recommended to use the --cpus and --memory options to limit resource usage.

Supported Hardware

Chip Series	Product Examples	Architecture
Ascend 910B	Atlas 800T A2, Atlas 900 A2 PoD	Auto-detected (ARM64/x86_64)
Ascend A3	Atlas 800T A3	Auto-detected (ARM64/x86_64)

Tips: Use the docker manifest inspect command to view the system architectures supported by an image.

License

View the license information for MindSpore and license informationincluded in these images.

As with all container images, pre-installed software packages (Python, system libraries, etc.) may be subject to their respective licenses.