unsloth: repin #144, dropping the commit upstream squashed as #28123 - #168
Merged
Conversation
|
You have reached your Codex usage limits for security reviews. Please try again later. |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Repins #144 from
586b15efto5a08a717.Why
586b15efhasc6e318e3("qwen4exp: cover the recurrent conv state for rollback") as an ancestor.Upstream took the same change from ServeurpersoCom as
ggml-org#28123, squashed to
0eadefeb, merged2026-09-01. A squash is not an ancestor of the pin, so the merge re-applies code the base already
has, exactly as the
_docblock in this file warns.Measured against three candidate bases:
586b15ef5a08a7170eadefeb)0f3a71beThe conflict is in
src/models/qwen4exp.cpp, and the hunk is our loop against the upstream one:same logic,
slot/n_slotsagainstt/K/s_slot. Upstream is already at b10759, so the nextnightly would have taken a base at or past b10731 and stopped there.
What changed on the branch
c6e318e3dropped; the branch also moves onto its declared basebase/upstream-662a0b012, which itwas sitting 6 commits behind. 7 commits become 6, no content edits.
Verification
Built the nightly-shaped artifact (current master + the new pin, merge clean) with CUDA:
test-llama-archs: qwen4exp OK on B200 (NMSE 8.47e-08) and CPU (0.00e+00), roundtrip OK. Same forqwen35, qwen3next and llama.
UD-IQ1_S+ the publishedmtp-...-Q8_0.ggufsidecar, 5 reps on one B200:Every MTP rep beats every off rep. The draft counters are identical in all five reps (61.7%
acceptance), and the off arm is asserted to draft nothing, so this is not an arm that silently ran
without a drafter. The off arm's own spread is wide (70.4 to 85.6) so treat the ratio as approximate;
the deterministic counters are the solid part.
dflashhas no registered arch-test config, so it prints no rows and is not covered here.