fix(qwen): enable Flash-Next multimodal processor paths - #7
Merged
Merged
Conversation
kanadaj
force-pushed
the
fix/qwen-multimodal-alias
branch
from
September 14, 2026 13:45
c8b4078 to
84ec70a
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The release checkpoint reports
model_type=qwen3_8_flash_next, but four equivalent Qwen VL processor paths are registered only for its development name,qwen4_exp. The release name therefore misses the established concurrent preprocessing policy, preprocessed-video metadata, timestamp-token construction, and image-only mRoPE fast path.This adds
qwen3_8_flash_nextto those four existing allowlists. It changes no algorithms, sampling settings, model weights, or behavior for other model types.Validation:
Known limitation: semantic short-video chronology remains unresolved. Earlier live image cases passed, but observed four-frame videos could still describe events in the wrong order. This PR restores the existing processor path; it does not claim to fix or qualify video ordering.
Validation used a read-only, network-disabled, GPU-disabled container. No image was built, published, or deployed.