This document is relevant for: Inf1, Inf2, Trn1, Trn2

Neuron Tools

Neuron provides debugging and profiling tools with the visualization support of the TensorBoard plugin. The Neuron helper tools assist in best practices for model onboarding and performance optimizations. The debugging and profiling tools provide monitoring of runtime and performance metrics insights.

About Neuron tools

Neuron tools are essential utilities that provide deep insights into model performance, system utilization, and hardware behavior on AWS Neuron devices. The toolset spans multiple categories to support the complete ML development lifecycle:

Profiling Tools: Advanced performance analysis capabilities including detailed execution traces, memory usage patterns, and optimization recommendations through Neuron Profiler and TensorBoard integration.

System Tools: Real-time monitoring and diagnostic utilities for tracking device utilization, process management, and hardware health across Neuron instances.

Performance Tools: Benchmarking and evaluation frameworks like NeuronPerf for measuring model performance, comparing configurations, and validating optimization strategies.

Helper Tools: Utility functions for model validation, system information gathering, and troubleshooting common deployment issues.

System Tools

• Neuron Monitor User Guide

• Neuron Top User Guide

• Neuron LS User Guide

• Neuron Sysfs User Guide

• NCCOM-TEST User Guide

• What’s New

TensorBoard Plugin for Neuron

• Neuron Plugin for TensorBoard (Trn1)

• Neuron Plugin for TensorBoard (Inf1)

• What’s New

Helper Tools

• Neuron Check Model

• Using Neuron GatherInfo Tool to collect debug and support information

Performance and Benchmarks Tools

• NeuronPerf (Beta)

• NCCOM-TEST User Guide

• Neuron Profile User Guide

• Neuron Profiler 2.0 (Beta) User Guide

Tutorials

TensorBoard

• Neuron Plugin for TensorBoard (Trn1)

• Track Training Progress in TensorBoard using PyTorch Neuron

• Profiling PyTorch Neuron (torch-neuronx) with TensorBoard

System Tools

• Track System Resource Utilization during Training with neuron-monitor using PyTorch Neuron

This document is relevant for: Inf1, Inf2, Trn1, Trn2