main
[Test refactor 1/5] Per-folder tests reorganization (#15725)
[generate] Completely stop relying on `cache_position` to prepare inputs (#44130)
[Generation] Add static ensemble verification for lossy speculative decoding (#45979)
Add distributed runtime utils and DistributedMixin (#47352)
:rotating_light: [`FA4`] Initial support (#42435)
Vectorize NoRepeatNGramLogitsProcessor and remove its host sync (#47571)
add paged attention tests support for XPU (#47163)
Fix stop string matching for byte-fragment tokens (#46530)
Use Python 3.9 syntax in tests (#37343)
[Fix] Fix multi-head latent attention (MLA) (#47761)