Vadim Gimpelson 52069012fe [Bugfix] Fix DeepGemm E8M0 accuracy degradation for Qwen3.5 FP8 on Blackwell (#38083) ...

Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>

2026-03-26 01:21:47 -07:00

..

kernels

[Bugfix][CI] Fix Marlin FP8 Linear Kernel for Compressed Tensors Format (#38092)

2026-03-25 21:11:43 -07:00

layers

[Bugfix] Fix DeepGemm E8M0 accuracy degradation for Qwen3.5 FP8 on Blackwell (#38083)

2026-03-26 01:21:47 -07:00

model_loader

[Performance] Auto-enable prefetch on NFS with RAM guard (#37673)

2026-03-24 17:31:14 -07:00

models

[Model] Add torch.compile support for InternVL vision encoder (#38049)

2026-03-25 23:52:29 -07:00

offloader

Bugfix for offloading+prefetch for GLM-4.7-FP8 (#37178)

2026-03-17 21:22:09 +08:00

warmup

[Bugfix] Fix AttributeError when serving MXFP8 models with DeepGEMM installed (#37358)

2026-03-19 17:58:33 +00:00

__init__.py

[Platform] Deprecate seed_everything (#31659)

2026-01-04 18:34:04 -08:00

custom_op.py

Add ability to replace oot ops when using lora (#37181)

2026-03-16 18:04:15 -07:00

parameter.py

[Mypy] Fix mypy for vllm/model_executor (except vllm/model_executor/layers) (#37904)

2026-03-24 17:14:01 +00:00

utils.py

[BugFix] Fix EPLB fail for MoeFP4 model with Marlin backend (#33262)

2026-01-29 16:52:11 +08:00