Qwen3: fix quality loss due to rope freq precision#3005
Merged
Conversation
greenrazer
force-pushed
the
qwen3_rope_fix
branch
from
915912b to
c9e19e9
Compare
Contributor
|
Thanks! |
john-sharratt
pushed a commit
to john-sharratt/candle
that referenced
this pull request
May 7, 2026
* qwen3 bugfix: compute rope freqs in f32 * qwen example: run model in bf16 on metal
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
When running Qwen3 at bf16, I noticed a substantial loss in quality and bugs (skipping tokens) especially when dealing with multi-piece tokens/multi-byte characters. Fixed by doing all intermediate rope freq and sin/cos computation in F32, then casting down to the requested dtype.
Also changed the candle example to run in bf16 by default on metal.