Skip to content

Repository navigation

fix(cpu-ep): stop handing attention and MoE nodes to ORT's CPU EP #1129

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
justinchuby merged 7 commits into from
Aug 17, 2026
Merged

fix(cpu-ep): stop handing attention and MoE nodes to ORT's CPU EP #1129

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
6d4063c
fix(cpu-ep): stop handing attention and MoE nodes to ORT's CPU EP
justinchuby Aug 17, 2026
dcf1e26
fix(cpu-ep): close three more silent declines found by review
justinchuby Aug 17, 2026
4264011
fix(cpu-ep): claim half-precision MoE, and stop claiming what we cann…
justinchuby Aug 17, 2026
e508e9e
fix(cpu-ep): mirror every attention/MoE factory limit at claim time
justinchuby Aug 17, 2026
34584d7
test(cpu-ep): make the PackedMHA mirror case actually reach the mirror
justinchuby Aug 17, 2026
c9a359b
Merge remote-tracking branch 'origin/main' into squad/leon-p5-no-decline
justinchuby Aug 17, 2026
2a704fc
Merge remote-tracking branch 'origin/main' into squad/leon-p5-no-decline
justinchuby Aug 17, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
480 changes: 480 additions & 0 deletions crates/onnx-runtime-ep-cpu-plugin/tests/fixtures/generate_attention_fixtures.py
View file
Open in desktop

Large diffs are not rendered by default.

226 changes: 226 additions & 0 deletions crates/onnx-runtime-ep-cpu-plugin/tests/fixtures/gqa_assignment_f32/model.onnx.textproto
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,226 @@
ir_version: 10
graph {
node {
input: "query"
input: "key"
input: "value"
input: "past_key"
input: "past_value"
input: "seqlens_k"
input: "total_sequence_length"
output: "output"
output: "present_key"
output: "present_value"
op_type: "GroupQueryAttention"
attribute {
name: "kv_num_heads"
i: 8
type: INT
}
attribute {
name: "num_heads"
i: 32
type: INT
}
domain: "com.microsoft"
}
name: "gqa_assignment_f32"
input {
name: "query"
type {
tensor_type {
elem_type: 1
shape {
dim {
dim_value: 1
}
dim {
dim_value: 1
}
dim {
dim_value: 4096
}
}
}
}
}
input {
name: "key"
type {
tensor_type {
elem_type: 1
shape {
dim {
dim_value: 1
}
dim {
dim_value: 1
}
dim {
dim_value: 1024
}
}
}
}
}
input {
name: "value"
type {
tensor_type {
elem_type: 1
shape {
dim {
dim_value: 1
}
dim {
dim_value: 1
}
dim {
dim_value: 1024
}
}
}
}
}
input {
name: "past_key"
type {
tensor_type {
elem_type: 1
shape {
dim {
dim_value: 1
}
dim {
dim_value: 8
}
dim {
dim_value: 1023
}
dim {
dim_value: 128
}
}
}
}
}
input {
name: "past_value"
type {
tensor_type {
elem_type: 1
shape {
dim {
dim_value: 1
}
dim {
dim_value: 8
}
dim {
dim_value: 1023
}
dim {
dim_value: 128
}
}
}
}
}
input {
name: "seqlens_k"
type {
tensor_type {
elem_type: 6
shape {
dim {
dim_value: 1
}
}
}
}
}
input {
name: "total_sequence_length"
type {
tensor_type {
elem_type: 6
shape {
dim {
dim_value: 1
}
}
}
}
}
output {
name: "output"
type {
tensor_type {
elem_type: 1
shape {
dim {
dim_value: 1
}
dim {
dim_value: 1
}
dim {
dim_value: 4096
}
}
}
}
}
output {
name: "present_key"
type {
tensor_type {
elem_type: 1
shape {
dim {
dim_value: 1
}
dim {
dim_value: 8
}
dim {
dim_value: 1024
}
dim {
dim_value: 128
}
}
}
}
}
output {
name: "present_value"
type {
tensor_type {
elem_type: 1
shape {
dim {
dim_value: 1
}
dim {
dim_value: 8
}
dim {
dim_value: 1024
}
dim {
dim_value: 128
}
}
}
}
}
}
opset_import {
domain: ""
version: 17
}
opset_import {
domain: "com.microsoft"
version: 1
}
Loading
Loading