- Introduction
- Installation
- Quick Start
- Docs
- Community
- Vulkan Vision
- Contributing
- Release Notes
- License
MindInsight provides MindSpore with easy-to-use debugging and tuning capabilities. During the training, data such as scalar, tensor, image, computational graph, model hyper parameter and training's execution time can be recorded in the file for viewing and analysis through the visual page of MindInsight.
Click to view the MindInsight design document, learn more about the design. Click to view the Tutorial documentation learn more about the MindInsight tutorial.
- The hardware platform supports Ascend, GPU and CPU.
- Confirm that Python 3.7.5 is installed.
- The versions of MindInsight and MindSpore must be consistent.
- If you use source code to compile and install, the following dependencies also need to be installed:
- All other dependencies are included in requirements.txt.
You can install MindInsight either by pip or by source code.
Install from PyPI:
pip install mindinsight
Install with customized version:
pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/{version}/MindInsight/any/mindinsight-{version}-py3-none-any.whl --trusted-host ms-release.obs.cn-north-4.myhuaweicloud.com -i https://pypi.tuna.tsinghua.edu.cn/simple
- When the network is connected, dependency items are automatically downloaded during .whl package installation. (For details about other dependency items, see requirements.txt). In other cases, you need to manually install dependency items.
{version}
denotes the version of MindInsight. For example, when you are downloading MindSpore 1.3.0,{version}
should be 1.3.0.- MindInsight supports only Linux distro with x86 architecture 64-bit or ARM architecture 64-bit.
git clone https://gitee.com/mindspore/mindinsight.git
You can choose any of the following installation methods:
-
Run the following command in the root directory of the source code:
cd mindinsight pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple python setup.py install
-
Build the
whl
package for installation.Enter the root directory of the source code, first execute the MindInsight compilation script in the
build
directory, and then execute the command to install thewhl
package generated in theoutput
directory.cd mindinsight bash build/build.sh pip install output/mindinsight-{version}-py3-none-any.whl -i https://pypi.tuna.tsinghua.edu.cn/simple
Execute the following command:
mindinsight start [--port PORT]
notes: the param --port default value is 8080
If it prompts the following information, the installation is successful:
Web address: https://127.0.0.1:8080
service start state: success
Before using MindInsight, the data in the training process should be recorded. When starting MindInsight, the directory of the saved data should be specified. After successful startup, the data can be viewed through the web page. Here is a brief introduction to recording training data, as well as starting and stopping MindInsight.
SummaryCollector is the interface MindSpore provides for a quick and easy collection of common data about computational graphs, loss values, learning rates, parameter weights, and so on. Below is an example of using SummaryCollector
for data collection, specifying the directory where the data is stored in ./summary_dir
.
...
from mindspore import SummaryCollector
summary_collector = SummaryCollector(summary_dir='./summary_dir')
model.train(epoch=1, ds_train, callbacks=[summary_collector])
For more ways to record visual data, see the MindInsight Tutorial.
After you've collected the data, when you launch MindInsight, specify the directory in which the data has been stored.
mindinsight start --summary-base-dir ./summary_dir [--port PORT]
notes: the param --port default value is 8080
After successful startup, visit https://127.0.0.1:8080
through the browser to view the web page.
Command of stopping the MindInsight service:
mindinsight stop [--port PORT]
notes: the param --port default value is 8080, you can stop the specified port MI service.
For more about MindInsight command_line,see the MindInsight Command_line.
More details about installation guide, tutorials and APIs, please see the User Documentation.
Check out how MindSpore Open Governance works.
- MindSpore Slack - Communication platform for developers.
- IRC channel at
#mindspore
(only for meeting minutes logging purpose) - Video Conferencing: TBD
- Mailing-list: https://mailweb.mindspore.cn/postorius/lists
Vulkan Vision(V-Vision) provides an unprecedented level of detail into the execution of Vulkan applications through dynamic instrumentation. V-Vision supports analyzing AI workloads implemented using the a compute pipeline as well as traditional raster and ray-tracing Vulkan applications. To use V-Vision please refer to the build instructions. For more information, please refer to the paper published at CGO 2021.
Welcome contributions. See our Contributor Wiki for more details.
The release notes, see our RELEASE.