Fix decode budget scaling - #165
Merged
Merged
Conversation
cunba-ai
pushed a commit
to cunba-ai/transcribe.cpp
that referenced
this pull request
Sep 21, 2026
Fix decode budget scaling (handy-computer#165). Conflict resolution in src/arch/qwen3_asr/model.cpp: our k_max_new=512 bump (be1f74f) is superseded by upstream's scaled decode budget — k_gen_reserve=256 acts as the input-gate reserve and decode-budget floor, and the actual budget is picked per run from the predicted transcript length (12 tokens/sec) clamped to the context ceiling. Took the upstream side verbatim.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Successor to #74 and #106
Should resolve #95 as well as some handy issues like #2087
In essence, previously we had some hard-coded constants which would truncate the output on some models, even if it could decode more. And this was basically an internal implementation trying to avoid degenerate decode loops, but we're going to fix that separately