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.