This document is relevant for: Inf1, Inf2, Trn1, Trn2, Trn3
Ask Amazon AI helper tools#
Use Kiro, Quick, and Amazon Q in the AWS console as your Neuron Experts for general Neuron technical guidance and to jumpstart your NKI kernel developement.
Guidelines for Quality Results#
Be Specific: Clearly state the task, desired output, and any constraints.
Provide Context: Mention specific versions, strategies, and any relevant performance requirements.
Request Complete Code: Ask for full implementations including imports, decorators, and main functions. Remember to always review and test the generated code before using it in production.
Ask for Explanations: Request comments or separate explanations for complex parts of the code.
Iterate: If the initial response isn’t satisfactory, refine your prompt based on the output. If you encounter issues or inaccuracies, consider rephrasing your prompt or breaking down complex tasks into smaller, more specific questions.
Fact check: Use Q as a starting point and supplement its output with official documentation, AWS NKI Samples repository, and your own expertise.
Example Prompts#
Note
Amazon AI helper tools may not be fully synched with the latest Neuron features. Therefore, they may not always produce optimal or fully accurate results.
“Explain the key features and benefits of AWS Neuron Kernel Interface (NKI).”
“How do different parallelism strategies (data, pipeline, tensor) affect training performance on Neuron?”
“What are the best practices for optimizing matrix multiplication operations using Neuron Kernel Interface (NKI)?”
“Provide complete Neuron Kernel Interface (NKI) code for a matrix multiplication kernel, including imports, decorators, and explanations of key optimizations. Focus on efficient tiling and data movement strategies.”
This document is relevant for: Inf1, Inf2, Trn1, Trn2, Trn3