KV handoff with DMA slicing APIs to avoid KV input/output copies.#1039

Open

quic-akuruvil wants to merge 28 commits into

quic:release/v1.22.0_tmpfrom

quic-akuruvil:dma_slice

Commits on May 25, 2026

rebase(transformers): align modeling wrappers, cache_utils and other changes to v5.5.4 and restore PyTorch/ORT parity (#876)

authored

Commits on May 26, 2026

fix: improve weight offloading to handle plain tensor attrs and use to_empty() (#952)
quic-rishinr
authored

Commits on May 27, 2026

fix retry timeout (#1012)
quic-mamta
authored

Commits on Jun 1, 2026

Change CCL input precision from int8 to int64 to align with compiler (#1003)
vjanfaza
authored

Commits on Jun 4, 2026

Commits on Jun 5, 2026

Commits on Jun 8, 2026

Commits on Jun 10, 2026

Added sample script for VLM disaggregated with KV slice
quic-akuruvil
committed