meta-pytorch/torchrec
A PyTorch extension library enabling massive-scale recommendation system training across distributed GPUs.
Not currently ranked — collecting fresh signals.
star history
TorchRec provides common sparsity and parallelism primitives for large-scale recommendation systems. It enables training and inference of models with embedding tables sharded across many GPUs using hybrid data-parallelism and model-parallelism. Key features include automated sharding planners, pipelined training, and optimized kernels via FBGEMM for powering production recommender systems at scale.
Frequently asked
- What is meta-pytorch/torchrec?
- A PyTorch extension library enabling massive-scale recommendation system training across distributed GPUs.
- Is torchrec open source?
- Yes — meta-pytorch/torchrec is open source, released under the BSD-3-Clause license.
- What language is torchrec written in?
- meta-pytorch/torchrec is primarily written in Python.
- How popular is torchrec?
- meta-pytorch/torchrec has 2.6k stars on GitHub.
- Where can I find torchrec?
- meta-pytorch/torchrec is on GitHub at https://github.com/meta-pytorch/torchrec.