MindInsight Documents

MindInsight is a visualized debugging and optimization tool of MindSpore. It visualizes the training process, model performance optimization, accuracy debugging, and inference result explanation. You can also use the command line provided by MindInsight to easily search for hyperparameters and migrate models. MindInsight helps you to easily obtain satisfactory model accuracy and performance.

MindInsight provides the following functions:

• Visualized training process (Collect Summary Record, View Dashboard)

• Training lineage and comparison

• Performance optimization

• Accuracy debugging

• Hyperparameter optimization

• Model migration

Using MindInsight to Visualize the Training Process

1. Collecting Data for Visualization

   Use SummaryCollector to record the training information in the training script and then perform the training.

2. Starting MindInsight for Visualization

   Start the MindInsight service and set the --summary-base-dir parameter to specify the directory for storing the summary log file.

3. Viewing Training Dashboard

   Open a browser, enter the MindInsight address in the address box, and click Training Dashboard to view details.

Using MindInsight to Analyze the Model Performance

1. Collecting Data for Analysis

   Call MindSpore Profiler APIs in the training script and then perform training.

2. Starting MindInsight for Analysis

   Start the MindInsight service and set the --summary-base-dir parameter to specify the directory for storing the performance data.

3. Analyzing Performance Data

   Open a browser, enter the MindInsight address in the address box, and click Profiling to view and analyze the training performance data.

Using MindInsight to Debug the Model Accuracy

1. Starting MindInsight in Debugger Mode

   Configure the --enable-debugger True --debugger-port 50051 parameter to start MindInsight in debugger mode.

2. 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.

3. Setting and Analyzing Watchpoints in MindInsight

   Open a browser, enter the MindInsight 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.

Using MindConverter to Migrate Models

1. Exporting the ONNX Model

   Export the ONNX model from the original script. For details, see the description provided by TensorFlow or PyTorch.

2. Migrating the Model

   Run the mindconverter command to migrate the model. After the migration, view the migration report and rectify the errors in the report.

3. Supplementing Scripts

   Based on the backbone model definition generated by MindConverter, supplement script code such as data processing, loss function, and training loop to form a complete script.

Installation

• MindInsight Installation

Guide

• Collecting Summary Record
• Viewing Dashboard
• Viewing Lineage and Scalars Comparison
• Use Mindoptimizer to Tune Hyperparameters
• Migrating From Third Party Frameworks With MindConverter
• Performance Profiling
• Debugger
• Explain Models
• Visualization of Training Optimization Process
• MindInsight Commands

Performance Tuning Guide

• Performance Tuning Guide

API References

• mindinsight.debugger
• mindconverter

References

• Overall Design of Training Visualization
• Computational Graph Visualization Design
• Tensor Visualization Design
• Performance Profiling Design
• FAQ