1 2 3[](https://pypi.org/project/mindspore) 4[](https://badge.fury.io/py/mindspore) 5[](https://pepy.tech/project/mindspore) 6[](https://hub.docker.com/r/mindspore/mindspore-cpu) 7[](https://github.com/mindspore-ai/mindspore/blob/master/LICENSE) 8[](https://join.slack.com/t/mindspore/shared_invite/zt-dgk65rli-3ex4xvS4wHX7UDmsQmfu8w) 9[](https://gitee.com/mindspore/mindspore/pulls) 10 11[查看中文](./README_CN.md) 12 13<!-- TOC --> 14 15- [What Is MindSpore](#what-is-mindspore) 16 - [Automatic Differentiation](#automatic-differentiation) 17 - [Automatic Parallel](#automatic-parallel) 18- [Installation](#installation) 19 - [Pip mode method installation](#pip-mode-method-installation) 20 - [Source code compilation installation](#source-code-compilation-installation) 21 - [Docker Image](#docker-image) 22- [Quickstart](#quickstart) 23- [Docs](#docs) 24- [Community](#community) 25 - [Governance](#governance) 26 - [Communication](#communication) 27- [Contributing](#contributing) 28- [Maintenance phases](#maintenance-phases) 29- [Maintenance status](#maintenance-status) 30- [Release Notes](#release-notes) 31- [License](#license) 32 33<!-- /TOC --> 34 35## What Is MindSpore 36 37MindSpore is a new open source deep learning training/inference framework that 38could be used for mobile, edge and cloud scenarios. MindSpore is designed to 39provide development experience with friendly design and efficient execution for 40the data scientists and algorithmic engineers, native support for Ascend AI 41processor, and software hardware co-optimization. At the meantime MindSpore as 42a global AI open source community, aims to further advance the development and 43enrichment of the AI software/hardware application ecosystem. 44 45<img src="https://gitee.com/mindspore/mindspore/raw/r1.5/docs/MindSpore-architecture.png" alt="MindSpore Architecture" width="600"/> 46 47For more details please check out our [Architecture Guide](https://www.mindspore.cn/docs/programming_guide/en/r1.5/architecture.html). 48 49### Automatic Differentiation 50 51There are currently three automatic differentiation techniques in mainstream deep learning frameworks: 52 53- **Conversion based on static compute graph**: Convert the network into a static data flow graph at compile time, then turn the chain rule into a data flow graph to implement automatic differentiation. 54- **Conversion based on dynamic compute graph**: Record the operation trajectory of the network during forward execution in an operator overloaded manner, then apply the chain rule to the dynamically generated data flow graph to implement automatic differentiation. 55- **Conversion based on source code**: This technology is evolving from the functional programming framework and performs automatic differential transformation on the intermediate expression (the expression form of the program during the compilation process) in the form of just-in-time compilation (JIT), supporting complex control flow scenarios, higher-order functions and closures. 56 57TensorFlow adopted static calculation diagrams in the early days, whereas PyTorch used dynamic calculation diagrams. Static maps can utilize static compilation technology to optimize network performance, however, building a network or debugging it is very complicated. The use of dynamic graphics is very convenient, but it is difficult to achieve extreme optimization in performance. 58 59But MindSpore finds another way, automatic differentiation based on source code conversion. On the one hand, it supports automatic differentiation of automatic control flow, so it is quite convenient to build models like PyTorch. On the other hand, MindSpore can perform static compilation optimization on neural networks to achieve great performance. 60 61<img src="https://gitee.com/mindspore/mindspore/raw/r1.5/docs/Automatic-differentiation.png" alt="Automatic Differentiation" width="600"/> 62 63The implementation of MindSpore automatic differentiation can be understood as the symbolic differentiation of the program itself. Because MindSpore IR is a functional intermediate expression, it has an intuitive correspondence with the composite function in basic algebra. The derivation formula of the composite function composed of arbitrary basic functions can be derived. Each primitive operation in MindSpore IR can correspond to the basic functions in basic algebra, which can build more complex flow control. 64 65### Automatic Parallel 66 67The goal of MindSpore automatic parallel is to build a training method that combines data parallelism, model parallelism, and hybrid parallelism. It can automatically select a least cost model splitting strategy to achieve automatic distributed parallel training. 68 69<img src="https://gitee.com/mindspore/mindspore/raw/r1.5/docs/Automatic-parallel.png" alt="Automatic Parallel" width="600"/> 70 71At present, MindSpore uses a fine-grained parallel strategy of splitting operators, that is, each operator in the figure is split into a cluster to complete parallel operations. The splitting strategy during this period may be very complicated, but as a developer advocating Pythonic, you don't need to care about the underlying implementation, as long as the top-level API compute is efficient. 72 73## Installation 74 75### Pip mode method installation 76 77MindSpore offers build options across multiple backends: 78 79| Hardware Platform | Operating System | Status | 80| :---------------- | :--------------- | :----- | 81| Ascend910 | Ubuntu-x86 | ✔️ | 82| | Ubuntu-aarch64 | ✔️ | 83| | EulerOS-aarch64 | ✔️ | 84| | CentOS-x86 | ✔️ | 85| | CentOS-aarch64 | ✔️ | 86| GPU CUDA 10.1 | Ubuntu-x86 | ✔️ | 87| CPU | Ubuntu-x86 | ✔️ | 88| | Ubuntu-aarch64 | ✔️ | 89| | Windows-x86 | ✔️ | 90 91For installation using `pip`, take `CPU` and `Ubuntu-x86` build version as an example: 92 931. Download whl from [MindSpore download page](https://www.mindspore.cn/versions/en), and install the package. 94 95 ```bash 96 pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/1.2.0-rc1/MindSpore/cpu/ubuntu_x86/mindspore-1.2.0rc1-cp37-cp37m-linux_x86_64.whl 97 ``` 98 992. Run the following command to verify the install. 100 101 ```python 102 import numpy as np 103 import mindspore.context as context 104 import mindspore.nn as nn 105 from mindspore import Tensor 106 from mindspore.ops import operations as P 107 108 context.set_context(mode=context.GRAPH_MODE, device_target="CPU") 109 110 class Mul(nn.Cell): 111 def __init__(self): 112 super(Mul, self).__init__() 113 self.mul = P.Mul() 114 115 def construct(self, x, y): 116 return self.mul(x, y) 117 118 x = Tensor(np.array([1.0, 2.0, 3.0]).astype(np.float32)) 119 y = Tensor(np.array([4.0, 5.0, 6.0]).astype(np.float32)) 120 121 mul = Mul() 122 print(mul(x, y)) 123 ``` 124 125 ```text 126 [ 4. 10. 18.] 127 ``` 128 129Use pip mode method to install MindSpore in different environments. Refer to the following documents. 130 131- [Using pip mode method to install MindSpore in Ascend environment](https://gitee.com/mindspore/docs/blob/master/install/mindspore_ascend_install_pip_en.md) 132- [Using pip mode method to install MindSpore in GPU environment](https://gitee.com/mindspore/docs/blob/master/install/mindspore_gpu_install_pip_en.md) 133- [Using pip mode method to install MindSpore in CPU environment](https://gitee.com/mindspore/docs/blob/master/install/mindspore_cpu_install_pip_en.md) 134 135### Source code compilation installation 136 137Use the source code compilation method to install MindSpore in different environments. Refer to the following documents. 138 139- [Using the source code compilation method to install MindSpore in Ascend environment](https://gitee.com/mindspore/docs/blob/master/install/mindspore_ascend_install_source_en.md) 140- [Using the source code compilation method to install MindSpore in GPU environment](https://gitee.com/mindspore/docs/blob/master/install/mindspore_gpu_install_source_en.md) 141- [Using the source code compilation method to install MindSpore in CPU environment](https://gitee.com/mindspore/docs/blob/master/install/mindspore_cpu_install_source_en.md) 142 143### Docker Image 144 145MindSpore docker image is hosted on [Docker Hub](https://hub.docker.com/r/mindspore), 146currently the containerized build options are supported as follows: 147 148| Hardware Platform | Docker Image Repository | Tag | Description | 149| :---------------- | :---------------------- | :-- | :---------- | 150| CPU | `mindspore/mindspore-cpu` | `x.y.z` | Production environment with pre-installed MindSpore `x.y.z` CPU release. | 151| | | `devel` | Development environment provided to build MindSpore (with `CPU` backend) from the source, refer to <https://www.mindspore.cn/install/en> for installation details. | 152| | | `runtime` | Runtime environment provided to install MindSpore binary package with `CPU` backend. | 153| GPU | `mindspore/mindspore-gpu` | `x.y.z` | Production environment with pre-installed MindSpore `x.y.z` GPU release. | 154| | | `devel` | Development environment provided to build MindSpore (with `GPU CUDA10.1` backend) from the source, refer to <https://www.mindspore.cn/install/en> for installation details. | 155| | | `runtime` | Runtime environment provided to install MindSpore binary package with `GPU CUDA10.1` backend. | 156 157> **NOTICE:** For GPU `devel` docker image, it's NOT suggested to directly install the whl package after building from the source, instead we strongly RECOMMEND you transfer and install the whl package inside GPU `runtime` docker image. 158 159- CPU 160 161 For `CPU` backend, you can directly pull and run the latest stable image using the below command: 162 163 ```bash 164 docker pull mindspore/mindspore-cpu:1.1.0 165 docker run -it mindspore/mindspore-cpu:1.1.0 /bin/bash 166 ``` 167 168- GPU 169 170 For `GPU` backend, please make sure the `nvidia-container-toolkit` has been installed in advance, here are some install guidelines for `Ubuntu` users: 171 172 ```bash 173 DISTRIBUTION=$(. /etc/os-release; echo $ID$VERSION_ID) 174 curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | apt-key add - 175 curl -s -L https://nvidia.github.io/nvidia-docker/$DISTRIBUTION/nvidia-docker.list | tee /etc/apt/sources.list.d/nvidia-docker.list 176 177 sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit nvidia-docker2 178 sudo systemctl restart docker 179 ``` 180 181 Then edit the file daemon.json: 182 183 ```bash 184 $ vim /etc/docker/daemon.json 185 { 186 "runtimes": { 187 "nvidia": { 188 "path": "nvidia-container-runtime", 189 "runtimeArgs": [] 190 } 191 } 192 } 193 ``` 194 195 Restart docker again: 196 197 ```bash 198 sudo systemctl daemon-reload 199 sudo systemctl restart docker 200 ``` 201 202 Then you can pull and run the latest stable image using the below command: 203 204 ```bash 205 docker pull mindspore/mindspore-gpu:1.1.0 206 docker run -it -v /dev/shm:/dev/shm --runtime=nvidia --privileged=true mindspore/mindspore-gpu:1.1.0 /bin/bash 207 ``` 208 209 To test if the docker image works, please execute the python code below and check the output: 210 211 ```python 212 import numpy as np 213 import mindspore.context as context 214 from mindspore import Tensor 215 from mindspore.ops import functional as F 216 217 context.set_context(mode=context.PYNATIVE_MODE, device_target="GPU") 218 219 x = Tensor(np.ones([1,3,3,4]).astype(np.float32)) 220 y = Tensor(np.ones([1,3,3,4]).astype(np.float32)) 221 print(F.tensor_add(x, y)) 222 ``` 223 224 ```text 225 [[[ 2. 2. 2. 2.], 226 [ 2. 2. 2. 2.], 227 [ 2. 2. 2. 2.]], 228 229 [[ 2. 2. 2. 2.], 230 [ 2. 2. 2. 2.], 231 [ 2. 2. 2. 2.]], 232 233 [[ 2. 2. 2. 2.], 234 [ 2. 2. 2. 2.], 235 [ 2. 2. 2. 2.]]] 236 ``` 237 238If you want to learn more about the building process of MindSpore docker images, 239please check out [docker](https://gitee.com/mindspore/mindspore/blob/r1.5/scripts/docker/README.md) repo for the details. 240 241## Quickstart 242 243See the [Quick Start](https://www.mindspore.cn/tutorials/en/r1.5/quick_start.html) 244to implement the image classification. 245 246## Docs 247 248More details about installation guide, tutorials and APIs, please see the 249[User Documentation](https://gitee.com/mindspore/docs). 250 251## Community 252 253### Governance 254 255Check out how MindSpore Open Governance [works](https://gitee.com/mindspore/community/blob/master/governance.md). 256 257### Communication 258 259- [MindSpore Slack](https://join.slack.com/t/mindspore/shared_invite/zt-dgk65rli-3ex4xvS4wHX7UDmsQmfu8w) - Communication platform for developers. 260- IRC channel at `#mindspore` (only for meeting minutes logging purpose) 261- Video Conferencing: TBD 262- Mailing-list: <https://mailweb.mindspore.cn/postorius/lists> 263 264## Contributing 265 266Welcome contributions. See our [Contributor Wiki](https://gitee.com/mindspore/mindspore/blob/master/CONTRIBUTING.md) for 267more details. 268 269## Maintenance phases 270 271Project stable branches will be in one of the following states: 272 273| **State** | **Time frame** | **Summary** | 274|-------------|---------------|--------------------------------------------------| 275| Planning | 1 - 3 months | Features are under planning. | 276| Development | 3 months | Features are under development. | 277| Maintained | 6 - 12 months | All bugfixes are appropriate. Releases produced. | 278| Unmaintained| 0 - 3 months | All bugfixes are appropriate. No Maintainers and No Releases produced. | 279| End Of Life (EOL) | N/A | Branch no longer accepting changes. | 280 281## Maintenance status 282 283| **Branch** | **Status** | **Initial Release Date** | **Next Phase** | **EOL Date**| 284|------------|--------------|--------------------------|----------------------------------------|-------------| 285| **r1.5** | Maintained | 2021-10-15 | Unmaintained <br> 2022-10-15 estimated | | 286| **r1.4** | Maintained | 2021-08-15 | Unmaintained <br> 2022-08-15 estimated | | 287| **r1.3** | Maintained | 2021-07-15 | Unmaintained <br> 2022-07-15 estimated | | 288| **r1.2** | Unmaintained | 2021-04-15 | End Of Life <br> 2022-04-15 estimated | | 289| **r1.1** | End Of Life | 2020-12-31 | | 2021-09-30 | 290| **r1.0** | End Of Life | 2020-09-24 | | 2021-07-30 | 291| **r0.7** | End Of Life | 2020-08-31 | | 2021-02-28 | 292| **r0.6** | End Of Life | 2020-07-31 | | 2020-12-30 | 293| **r0.5** | End Of Life | 2020-06-30 | | 2021-06-30 | 294| **r0.3** | End Of Life | 2020-05-31 | | 2020-09-30 | 295| **r0.2** | End Of Life | 2020-04-30 | | 2020-08-31 | 296| **r0.1** | End Of Life | 2020-03-28 | | 2020-06-30 | 297 298## Release Notes 299 300The release notes, see our [RELEASE](https://gitee.com/mindspore/mindspore/blob/master/RELEASE.md). 301 302## License 303 304[Apache License 2.0](https://gitee.com/mindspore/mindspore#/mindspore/mindspore/blob/master/LICENSE) 305