skip target model mm emb computation when draft is text-only#1

Open

kkt-cohere wants to merge 102 commits intomainfrom

Commits on Jan 29, 2026

skip target model mm emb computation when draft is text-only
kkt-cohere
committed
[Bugfix] Fix broken GLM-OCR initialization (#33350)
Isotr0py
authored
[ez] Delete torch25_custom_graph_pass (#33287)
angelayi
authored
[Chore] Move `MediaConnector` to `vllm.multimodal.media` (#33324)
DarkLight1337
authored
[BUGFIX][XPU] fix memory check after XPU reuse GPU_worker (#33358)
xuechendi
authored
[NVIDIA] [feat] Integrate flashinfer Trtllmgen bf16 moe (#32954)
Linda-Stadter
authored
[Bugfix][Kernel] Fix negative memory offset in GDN Triton kernel (#33326)
CarstyYou
authored
Add Triton fused MoE config for B200 (Nemotron Nano) (#32804)
danisereb
authored
[release] Minor fixes to release annotation and wheel upload (#33129)
khluu
authored
[Bugfix] Enable Triton MoE for FP8 per-tensor dynamic (#33300)
mgoin
authored
[Docs] Adding links and intro to Speculators and LLM Compressor (#32849)

aireilly
and
hmellor
authored

Commits on Jan 30, 2026

Commits on Jan 31, 2026

Commits on Feb 1, 2026