vllm
[Multimodal][Core] Optimize multimodal preprocessing cache by hashing image bytes instead of pixel values
#29621
Merged
Merged
[Multimodal][Core] Optimize multimodal preprocessing cache by hashing image bytes instead of pixel values #29621
DarkLight1337
merged
10 commits into
vllm-project:main
from
ImaGoodFella:image_bytes_caching
gemini-code-assist
commented
on 2025-11-27
DarkLight1337
commented
on 2025-11-27
DarkLight1337
commented
on 2025-11-28
DarkLight1337
commented
on 2025-12-01
DarkLight1337
commented
on 2025-12-01
DarkLight1337
approved these changes
on 2025-12-01
disabled auto-merge 74 days ago
Head branch was pushed to by a user without write access
Head branch was pushed to by a user without write access
disabled auto-merge 74 days ago
Head branch was pushed to by a user without write access
Head branch was pushed to by a user without write access
updated hash to use image_bytes instead of pixel_values
21a97020
fix formatting
877f9638
added wrapper for bytes
89ae13b9
fixed typing
4fdaac46
removed unnecessary metadata
c7d625d3
renamed to get_item_for_hash
32f6ce32
added __array__ method to deal with the casting issues
dedb0572
fixed base
057cb258
Merge branch 'main' into image_bytes_caching
24565a9d
fixed tests
65638533
disabled auto-merge 73 days ago
Head branch was pushed to by a user without write access
Head branch was pushed to by a user without write access
Reviewers
Assignees
No one assigned
Labels
performance
ready
multi-modality
Milestone
No milestone
Login to write a write a comment.
Login via GitHub