This project demonstrates the process of fine-tuning and compiling an SDXL model using AWS Neuron architecture. The compilation was carried out on an inf2 EC2 instance with the Hugging Face AMI to optimize the model for AWS Neuron's hardware.
The notebook included in this repository provides detailed steps for others looking to compile their own models using a similar setup. It offers valuable insights into configuring, fine-tuning, and optimizing models on AWS infrastructure for improved performance and scalability.
Feel free to explore and adapt the notebook to suit your specific use case.
In addition to the compilation notebook, two more notebooks are provided:
-
Text-to-Image Inference Notebook: This notebook demonstrates how to inference the compiled SDXL model using the text-to-image method in an AWS SageMaker notebook environment.
-
Image-to-Image Inference Notebook: This notebook showcases the inference process for the compiled SDXL model using the image-to-image method in AWS SageMaker.
These resources are available for reference and adaptation to suit various use cases involving AWS Neuron and SageMaker.