NVIDIA/recsys-examples v25.10

v25.10

Repository: NVIDIA/recsys-examples

Tag: v25.10

Published: 2025-11-11T02:01:57Z

Prerelease: no

Release notes:

What's Changed

Features & Enhancements

• Add sequence parallelism by @JacoCheung in https://github.com/NVIDIA/recsys-examples/pull/216
• Decouple scaling seqlen from max_seqlen in hstu attn by @geoffreyQiu in https://github.com/NVIDIA/recsys-examples/pull/208
• Fea support lru score dump load by @shijieliu in https://github.com/NVIDIA/recsys-examples/pull/186
• Gradient clipping by reusing TorchRec&FBGEMM's parameters by @jiashuy in https://github.com/NVIDIA/recsys-examples/pull/223
• [HSTU]Add SM 89 support by @JacoCheung in https://github.com/NVIDIA/recsys-examples/pull/217
• allow allow_overwrite in DynamicEmbDump by @fshhr46 in https://github.com/NVIDIA/recsys-examples/pull/206

But Fixs

• Fix LFU mode frequency count bug by @z52527 in https://github.com/NVIDIA/recsys-examples/pull/176
• Fix config bug when using torchrec's STBE in benchmark by @jiashuy in https://github.com/NVIDIA/recsys-examples/pull/193
• Fix IMA in incremental dump and test the dumped embeddings by @jiashuy in https://github.com/NVIDIA/recsys-examples/pull/211
• Fix rab num heads by @JacoCheung in https://github.com/NVIDIA/recsys-examples/pull/222
• Fix IMA caused by wrong worker id for device of which max threads is … by @jiashuy in https://github.com/NVIDIA/recsys-examples/pull/220

Misc

• Code reorganization for hstu training and inference by @geoffreyQiu in https://github.com/NVIDIA/recsys-examples/pull/202
• Add embedding pooling kernel by @z52527 in https://github.com/NVIDIA/recsys-examples/pull/215

Full Changelog: https://github.com/NVIDIA/recsys-examples/compare/v25.09...v25.10