Unify canonical MoE folding across Qwen and GLM - #44
Closed
kiddyboots216 wants to merge 17 commits into
Closed
Conversation
Broly Security ScanNote ✅ Clean scan Note Re-scan this PR anytime with
|
kiddyboots216
force-pushed
the
feature/canonical-moe-fold-unification-20260810
branch
from
August 11, 2026 18:01
8285fde to
e08c0a4
Compare
kiddyboots216
changed the base branch from
main
to
feature/qwen3-dense-exact-kernels-20260810
August 11, 2026 18:02
This was referenced Aug 11, 2026
Closed
kiddyboots216
force-pushed
the
feature/qwen3-dense-exact-kernels-20260810
branch
from
August 12, 2026 02:05
61d8222 to
f123869
Compare
kiddyboots216
force-pushed
the
feature/canonical-moe-fold-unification-20260810
branch
from
August 12, 2026 02:29
e08c0a4 to
5bb7a3f
Compare
kiddyboots216
force-pushed
the
feature/canonical-moe-fold-unification-20260810
branch
from
August 12, 2026 03:15
5bb7a3f to
0f41afe
Compare
kiddyboots216
force-pushed
the
feature/canonical-moe-fold-unification-20260810
branch
from
August 12, 2026 03:19
0f41afe to
65ae56a
Compare
kiddyboots216
force-pushed
the
feature/canonical-moe-fold-unification-20260810
branch
from
August 12, 2026 04:02
65ae56a to
ecfe560
Compare
kiddyboots216
added a commit
that referenced
this pull request
Aug 12, 2026
The Qwen3.5/GLM exact lanes reduce MoE partials with the canonical adjacent-pair BF16 fold (#44/#16); DSV4 reproduces the pinned NCCL-tree contributor order [1..7, 0]. These are different byte programs, so the DSV4 combine (and the variable-row gather primitives it needs) now lives in dsv4_native_combine.py with the order baked in and the unification hazard documented; ep_native_combine.py and its tests return to main's content, removing the #44 -> #45 textual conflict entirely. Pure code motion: the 4-decision base ruler replay was re-verified byte-equal (K3 = 0 x4) after the relocation.
Contributor
Author
|
Superseded by #57, which consolidates the complete exact trainer-serving stack directly onto main. |
kiddyboots216
deleted the
feature/canonical-moe-fold-unification-20260810
branch
August 17, 2026 23:41
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
canonical_moe_fold_v1as the versioned logical adjacent-pair BF16 tree.DSV4 consumes this primitive through #45. The former architecture-scoped combine divergence is gone; DSV4 retains only its variable-row transport before the shared fold.
Stack
10439526fb35c8858bb3ebc8c382172cb9ef7f24c9e9dd662200bc576fffb87a3c946f5eaecaa29d5bb7a3f37e6f08d9c3299b69264a2ad24f7846b8Validation