MindSpore Insight Documents
MindSpore Insight is a visualized debugging and optimization tool, which helps users achieve better model precision and performance.
MindSpore Insight visualizes the training process, model performance optimization, and accuracy debugging. You can also use the command line provided by MindSpore Insight to easily search for hyperparameters and migrate models.
MindSpore Insight provides the following functions:
Visualized training process (Collect Summary Record, View Dashboard)
Code repository address: <https://gitee.com/mindspore/mindinsight>
Using MindSpore Insight to Visualize the Training Process
Collecting Data for Visualization
Use SummaryCollector to record the training information in the training script and then perform the training.
Starting MindSpore Insight for Visualization
Start the MindSpore Insight service and set the
--summary-base-dir
parameter to specify the directory for storing the summary log file.-
Open a browser, enter the MindSpore Insight address in the address box, and click Training Dashboard to view details.
Using MindSpore Insight to Analyze the Model Performance
-
Call MindSpore Profiler APIs in the training script and then perform training.
Starting MindSpore Insight for Analysis
Start the MindSpore Insight service and set the
--summary-base-dir
parameter to specify the directory for storing the performance data.-
Open a browser, enter the MindSpore Insight address in the address box, and click Profiling to view and analyze the training performance data.
Using MindSpore Insight to Debug the Model Accuracy
Starting MindSpore Insight in Debugger Mode
Configure the
--enable-debugger True
--debugger-port 50051
parameter to start MindSpore Insight in debugger mode.Running the Training Script in Debugger Mode
Set the environment variable
export ENABLE_MS_DEBUGGER
to True to specify the debugger mode for training. Set the debugging service and port to be connected for training:export MS_DEBUGGER_HOST=127.0.0.1
.export MS_DEBUGGER_PORT=50051
.Run the training script.Setting and Analyzing Watchpoints in MindSpore Insight
Open a browser, enter the MindSpore Insight address in the address box, click the Debugger tab page, set the watchpoints after the training is connected, and analyze the data such as the computational graphs, tensors, and watchpoint hits to identify the root cause of the accuracy problem.