Shypd.aiThe largest open index of AI toolsTorchRec is an Open Source & Models tool that provides a PyTorch domain library for recommendation systems. It offers common sparsity and parallelism primitives for building and training large-scale recommendation models, powering production RecSys models at Meta.
Trending
About
TorchRec is a PyTorch domain library specifically designed for large-scale recommendation systems (RecSys). It provides essential sparsity and parallelism primitives, enabling the training and inference of models with extensive embedding tables sharded across multiple GPUs. This library is crucial for Meta's production RecSys models and has been used to accelerate advancements in the field, including the latest version of Meta's DLRM. Key features include various sharding strategies, an automatic sharding plan planner, pipelined training for performance, optimized kernels powered by FBGEMM, and quantization support. It also offers common RecSys modules and datasets, along with end-to-end training examples.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
