vllm
[Models][Qwen3VL] Speedup `fast_pos_embed_interpolate`
#26647

Merged

[Models][Qwen3VL] Speedup `fast_pos_embed_interpolate` #26647

Isotr0py merged 3 commits into vllm-project:main from lgeiger:qwen3-pos-interpolate

Simplify and speedup `fast_pos_embed_interpolate`

18cfb2a0

Speedup `fast_pos_embed_interpolate` by using more vectorized ops

95103bbf

lgeiger requested a review from

sighingnow 62 days ago

mergify added qwen

lgeiger commented on 2025-10-11

gemini-code-assist commented on 2025-10-11

ywang96 approved these changes on 2025-10-12

DarkLight1337 enabled auto-merge (squash) 61 days ago

github-actions added ready

Simplify tiling

24b6717d

disabled auto-merge 61 days ago
Head branch was pushed to by a user without write access

Isotr0py merged a6049be7 into main 61 days ago

lgeiger deleted the qwen3-pos-interpolate branch 61 days ago

Reviewers

ywang96

gemini-code-assist

sighingnow

Assignees

No one assigned

Labels

ready qwen

Milestone

No milestone