With hundreds of millions of customers shopping daily on Amazon, it’s critical they help them discover the right product from their massive catalog of products.
Amazon’s Deep Scalable Sparse Tensor Network Engine (DSSTNE or “Destiny”) is a deep learning framework built from the ground up to help researchers develop search and recommendation systems. With multi-GPU support, DSSTNE can automatically distribute computational workloads across all available GPUs, speeding up training of larger models without a lot of effort. As a result, DSSTNE can be used to build recommendations systems that model ten million unique products instead of being limited to hundreds of thousands possible with other solutions.
The DSSTNE roadmap includes support for the types of convolutional layers used in image processing, and additional recurrent layers needed for many natural language understanding and speech recognition tasks.
Find on GitHub >>
Amazon Releases Open-Source Deep Learning Software
May 11, 2016
Discuss (0)
Related resources
- GTC session: Which GPU for Open Source LLM Applications? Best Practices on Hyperstack (Presented by NexGen Cloud)
- GTC session: Bringing Generative AI and Vision AI to Production at the Edge with Metropolis Microservices for Jetson
- GTC session: Taking Generative AI and Vision AI to Production at the Edge
- NGC Containers: ASR Parakeet CTC Riva 1.1b
- NGC Containers: MATLAB
- NGC Containers: TensorRT October 2024 (PB 24h2)