Add MiniMax-H3 Ref2VA Turbo adapter support - #295
Merged
Conversation
a-github-name
marked this pull request as ready for review
August 14, 2026 09:57
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Why
Mere.run previously treated every MiniMax-H3 adapter as FL2VA-only and rejected adapters whenever ordered Ref2VA references were present. LightX2V now publishes a dedicated Ref2VA Turbo adapter, so the runtime needs to select the Ref2VA base, require references, and fuse the adapter into resident BF16 weights without changing the existing compact checkpoint or compiled denoising path.
Runtime behavior
The new managed adapter is:
It is pinned to upstream revision
5d1d4829fe614c1b93fcfd9cc7718e9ba71f73e1, exact size1,383,677,768bytes, and SHA-2569e642fc8749c74f8da5e2382877ab5c7aa37b9a73b7fd0d6d457bd1b3cb1ae99.Ref2VA Turbo requires at least one ordered image or video reference and resident BF16 transformer weights. Explicit
--h3-weight-mode quantizedrequests fail preflight with a focused diagnostic. The adapter selects the published four-evaluation schedule, video/audio shifts 12/3, and LightX2V alpha 8.Validation
origin/mainat395088ee./scripts/check.shadapter pulldownloaded and independently re-verified the exact byte count and SHA-256Licensing
The adapter is Apache-2.0. The MiniMax-H3 base-model Community License and its geographic/use restrictions remain unchanged.