MOVZXandHinata e7cf27cc6f [Model] Fix runtime abliteration for Qwen3.5 models and Qwen3-Next MTP ...

Qwen3_5Model.__init__ skips Qwen3NextModel.__init__, so the abliteration
projector was never installed and embedding lookup raised AttributeError
when VLLM_ABLITERATION was set. Install it in Qwen3_5Model. Project the
MTP token embeddings of Qwen3NextMultiTokenPredictor too, matching the
abliterated releases, which edit the MTP embed_tokens rows.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>

2026-10-14 08:12:14 +00:00

..

determinism

[Bugfix][Determinism] Preserve output dtype in batch-invariant mean (#59106)

2026-10-04 17:07:46 +00:00

hw_agnostic

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

kernels

[Fusion] Support 3D activation quantization fusion and enable it for Pixtral (#59612)

2026-10-09 15:12:03 +08:00

layers

Merge branch 'main' into patches

2026-10-09 17:41:13 +00:00

model_loader

Remove Python 3.10 support now that it is EOL (#60402)

2026-10-08 02:30:52 -07:00

models

[Model] Fix runtime abliteration for Qwen3.5 models and Qwen3-Next MTP

2026-10-14 08:12:14 +00:00

offloader

[MRV2] Release weight offloader on shutdown (#57834)

2026-09-22 15:01:20 -04:00

warmup

Merge branch 'main' into patches

2026-10-09 17:41:13 +00:00

__init__.py

…

custom_op.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

parameter.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

utils.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00