Support new arguments for exllamav3 Qwen3.5/3.6 MTP#421
Draft
alexhunt7 wants to merge 2 commits into
Draft
Conversation
Adds draft_arch_override and num_draft_tokens to DraftModelConfig so Qwen3.5/3.6 BF16 directories can be loaded as MTP-only draft models (arch_override="Qwen3_5MTPDraftModel"). Threads both options through to Config.from_directory and AsyncGenerator. If draft_arch_override is set but draft_model_name is omitted, treat the main model_directory as the source for the draft model. This covers the case where the same checkpoint contains both the trunk and the mtp.* tensors — no need to point at a separate directory or extract the MTP head into its own dir.
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Is your pull request related to a problem? Please describe.
Support new arguments for exllamav3 Qwen3.5/3.6 MTP support.
Also adds draft acceptance metrics to the logs (in a separate commit).
Why should this feature be added?
MTP speeds things up tremendously. This supports adds support for configuring MTP in conjunction with turboderp-org/exllamav3#206
Examples
~2x token generation speed in Qwen3.6 27B on my 5070ti + 3060ti.
Additional context
Keeping as a draft until turboderp-org/exllamav3#206 is merged, but this is otherwise ready to go.