Back to News
PyTorch 2.15 Rebuilds Distributed Training and Cuts Decoding Overhead
The new release flips the rewritten NCCL backend on by default, adds one-sided transports, and accelerates NVFP4 decode on Blackwell.
Source
PyTorch
Published
Author
AlphaSignal Newsroom
Read
1 min read
The new release flips the rewritten NCCL backend on by default, adds one-sided transports, and accelerates NVFP4 decode on Blackwell.
Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.
Read original report
