IFU release v2.6#406

Merged

wangye805 merged 31 commits intorelease_v2.6_rocmfrom

yewang12/ifu_release_v2.6_rocm

Feb 19, 2026

Commits on Jul 20, 2025

Changed VERSION to 2.6.0
KshitijLakhani
committed

Commits on Jul 22, 2025

[PyTorch] Remove GH pinned deps (#1961)

ksivaman
authored and
KshitijLakhani
committed
[PyTorch] Reset FP8 weight workspace if usages are invalid (#1972)

timmoon10
authored and
KshitijLakhani
committed
Fix the condition error when checking fp8 attn in `get_attention_backend` (#1965)

authored and
KshitijLakhani
committed
[Common] Skip cuDNN 9.10.0/9.10.1 due to bugs (#1937)

authored and
KshitijLakhani
committed
[PyTorch] Debug linear layer when saving original input and using debug quantizer (#1963)

authored and
KshitijLakhani
committed
[Common] Improved performance of mxfp8 cast kernels (#1628)

authored and
KshitijLakhani
committed

Commits on Jul 24, 2025

Commits on Jul 25, 2025

Commits on Jul 29, 2025

Fix the use-after-free bug in unfused normalization (#2002)

ptrendx
authored and
KshitijLakhani
committed

Commits on Jan 3, 2026

Merge remote-tracking branch 'upstream/release_v2.6' into yewang12/ifu_release_v2.6_rocm
wangye805
committed

Commits on Jan 9, 2026

[ROCm] Resolve conflicts
wangye805
committed

Commits on Jan 31, 2026

[ROCm] address reviewer comments
wangye805
committed

Commits on Feb 1, 2026

CI: Serialize core sgpu test (#426)

leo-automation
authored and
wangye805
committed

Commits on Feb 2, 2026

Commits on Feb 3, 2026

Commits on Feb 13, 2026

Bump up FA to 2.8.3 (#2282)

committed

Commits on Feb 17, 2026

[CI] Use default checkout depth (#438)

matthiasdiener
authored and
wangye805
committed
CI: Update runners (#445)

leo-automation
authored and
wangye805
committed

Commits on Feb 18, 2026

[ROCm] skip pytorch distributed test_sanity due to a flaky failure
wangye805
committed