AWS Silicon Migration Assessment
By:
Loka
If you've been utilizing NVIDIA GPUs for ML workloads but are intrigued by the potential cost and performance benefits of AWS Silicon, Loka provides a simple path to evaluation. Our AWS Silicon Migration Assessment benchmarks your existing ML workflows on NVIDIA GPUs against AWS silicon. This includes converting models from CUDA to Neuron and assessing performance on AWS Trainium instances, offering a direct comparison, benchmarking report and roadmap to migration.
Overview
Here's how it works:
Phase 1 - Discovery (1 week): Review your NVIDIA GPU-based ML setup, focusing on algorithms, training loops, and datasets. Establish benchmarks and success metrics compared to your current CUDA models.
Phase 2 - Neuron Porting & Model Deployment (4-8 weeks): Our team ports your ML models to AWS's Neuron and deploys them on Trainium instances, aiming for an isolated performance benchmark against NVIDIA GPUs.
Phase 3 - Benchmarking (1-2 weeks): Conduct a detailed performance and cost comparison between your Neuron models on Trainium and the NVIDIA GPU models. A conclusive report outlines benchmark reporting, comparisons and migration strategies to AWS silicon.
| Sold by | Loka |
| Categories | |
| Fulfillment method | Professional Services |
Pricing Information
This service is priced based on the scope of your request. Please contact seller for pricing details.
Support
To schedule your free Consultation contact Loka at AWS-team@loka.com, or your AWS representative.