Repository navigation
[DO NOT MERGE] Adding qmoe op support for qnn-gpu - #819
Open
sachjang-qti wants to merge 3 commits into
Open
sachjang-qti wants to merge 3 commits into
sachjang-qti wants to merge 3 commits into
Conversation
sachjang-qti
requested review from
minfhong-qti,
qti-ashwshan,
qti-chuteng,
qti-shubham,
qti-yuduo,
tirupath-qti and
yuhuchua-qti
as code owners
September 11, 2026 13:21
sachjang-qti
changed the title
sachjang-qti
changed the title
sachjang-qti
force-pushed
the
dev/sachjang-qti/qmoe-op-support-qnn-gpu
branch
2 times, most recently
from
September 16, 2026 08:35
db2a20c to
5c7e2ca
Compare
vjatoth-qti
reviewed
Sep 28, 2026
| OrtNodeAttrHelper h(node_unit); | ||
| const uint32_t block_size = static_cast<uint32_t>(h.Get("block_size", int64_t(32))); | ||
| const uint32_t bits = static_cast<uint32_t>(h.Get("expert_weight_bits", int64_t(4))); | ||
| const size_t map[14] = {0, 1, 2, 5, 3, 4, 6, 7, 8, 9, 10, 11, 12, 13}; |
Collaborator
There was a problem hiding this comment.
Here we are processing only 14 input slots. QMoE also has optional input slot 14, router_weights. IsOpSupported does not reject it, so models providing router weights may be assigned to QNN while the input is dropped, changing routing behavior.
sachjang-qti
force-pushed
the
dev/sachjang-qti/qmoe-op-support-qnn-gpu
branch
from
October 7, 2026 07:29
5c7e2ca to
dbb57f1
Compare
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
Adding QMoE OP support for qnn-gpu
Motivation and Context
Required to run MoE OP on qnn-gpu needed by gpt-oss and other upcomming models