Repository navigation
fix(vlm): load frozen vision weights and disable unused MTP - #4315
Draft
ehosseiniasl wants to merge 1 commit into
Draft
ehosseiniasl wants to merge 1 commit into
ehosseiniasl wants to merge 1 commit into
Conversation
ehosseiniasl
force-pushed
the
ehsan/main-video-media-alignment-vllm029
branch
from
October 1, 2026 00:43
d9095dc to
c453c2b
Compare
ehosseiniasl
force-pushed
the
ehsan/main-video-frozen-vision-vllm029
branch
from
October 1, 2026 00:43
5e7d39f to
640d554
Compare
Signed-off-by: ehosseiniasl <ehsan.hosseiniasl@gmail.com>
ehosseiniasl
force-pushed
the
ehsan/main-video-media-alignment-vllm029
branch
from
October 1, 2026 06:48
c453c2b to
9d6ac0a
Compare
ehosseiniasl
force-pushed
the
ehsan/main-video-frozen-vision-vllm029
branch
from
October 1, 2026 06:48
640d554 to
2eb476a
Compare
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Preserves explicit vLLM
load_format: autowhen frozen vision weights are excluded from trainer-to-vLLM refit, and removes inherited repeated-layer/pattern state whenmtp_num_layers: 0. This is PR 5 of the stacked current-main video/tracking port and contains one signed commit.Series
The series is intentionally stacked so each review contains one focused commit. Merge or retarget in numeric order. The final cached-video source-timing fix is #4310, adapted from #4309.
Validation