onnxruntime
[CUDA][WebGPU] Add packed sparse attention indexer for continuous batching
#32618
Merged
Overview
Commits
103
Changes
View On GitHub

[CUDA][WebGPU] Add packed sparse attention indexer for continuous batching #32618

kunal-vaishnavi merged 103 commits into main from copilot/copilotimplement-packed-sparse-attention-indexer
copilot-swe-agent
Copilot Implement SparseAttentionIndexer CUDA operator
3e309f4f
Copilot Harden SparseAttentionIndexer validation and tests
d0a8e7a5
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
cb58b7e4
Copilot Implement SparseAttentionIndexer for WebGPU
9acfe238
Copilot Address SparseAttentionIndexer review feedback
7adf1398
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
f89ea1ec
Copilot Implement SparseAttentionIndexer CUDA operator
d96d08f7
Copilot Harden SparseAttentionIndexer validation and tests
b5e79853
Copilot Address SparseAttentionIndexer review feedback
213b7316
kunal-vaishnavi Update onnxruntime/test/python/onnxruntime_test_python_symbolic_shape…
f3302871
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
765eadcd
Copilot Fix WebGPU SparseAttentionIndexer review issues
f74c5513
Copilot Fix SparseAttentionIndexer CI failures
c24495d5
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
38c341ce
Copilot Fix SparseAttentionIndexer CI failures
e3ea06a0
Copilot Implement SparseAttentionIndexer CUDA operator
814f5bf8
Copilot Harden SparseAttentionIndexer validation and tests
580d9f05
Copilot Address SparseAttentionIndexer review feedback
d1657581
kunal-vaishnavi Update onnxruntime/test/python/onnxruntime_test_python_symbolic_shape…
c695c8e8
Copilot Fix SparseAttentionIndexer CI failures
ea75f956
Copilot Update generated contrib operator documentation
2b06f06a
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
b028e65d
Copilot Gate FP8 XQA test on supported GPUs
59bc669e
Copilot Address SparseAttentionIndexer review feedback
aaa82997
Copilot Gate XQA dispatch tests by CUDA architecture
a0e9d1ba
Copilot Merge remote-tracking branch 'origin/main' into copilot/implement-spa…
ad9af07b
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
4df139cc
Copilot Initial plan
dbaeb30d
copilot-swe-agent copilot-swe-agent assigned copilot-swe-agent copilot-swe-agent 41 days ago
copilot-swe-agent copilot-swe-agent assigned kunal-vaishnavi kunal-vaishnavi 41 days ago
azure-pipelines
copilot-swe-agent copilot-swe-agent requested a review from kunal-vaishnavi kunal-vaishnavi 41 days ago
Copilot Support shared SparseAttentionIndexer caches
f039ff6d
Copilot Implement packed sparse attention indexer
cf546932
Copilot Harden packed indexer state validation
6f8041c0
Copilot Validate packed indexer capacity bounds
5b9d10eb
Copilot Fix packed indexer output-count test
de0e5874
Copilot Consolidate SparseAttentionIndexer cache state
5a37cb63
kunal-vaishnavi kunal-vaishnavi marked this pull request as ready for review 41 days ago
kunal-vaishnavi kunal-vaishnavi requested a review from copilot-pull-request-reviewer copilot-pull-request-reviewer 41 days ago
kunal-vaishnavi kunal-vaishnavi changed the title [WIP] Add packed sparse attention indexer contrib operator [CUDA][WebGPU] Add packed sparse attention indexer for continuous batching 41 days ago
copilot-pull-request-reviewer
copilot-pull-request-reviewer commented on 2026-09-16
Copilot Fix SparseAttentionIndexer test narrowing warning
76d6532b
Copilot Update generated SparseAttentionIndexer docs
daf73e35
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
2652363c
Copilot Adapt WebGPU indexer to shared cache ABI
3d328b27
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
2125b41b
Copilot Address packed indexer review feedback
df262844
Copilot Fix packed indexer CI failures
8ece210c
Copilot Fix packed indexer schema validation
d0b6fafa
Copilot Reduce packed indexer WebGPU bindings
e81906b1
kunal-vaishnavi Fuse query normalization into SparseAttentionIndexer
c6efe240
kunal-vaishnavi Merge branch 'stack-32526' into stack-32528
6a13c37a
kunal-vaishnavi Fuse query normalization in WebGPU sparse indexer
fe9d0594
kunal-vaishnavi Merge branch 'stack-32528' into stack-32618
a73ff17b
kunal-vaishnavi Fuse query normalization into PackedSparseAttentionIndexer
f32b9bff
kunal-vaishnavi Merge remote-tracking branch 'origin/main' into stack-32526
efa7f564
kunal-vaishnavi Merge branch 'stack-32526' into stack-32528
72d02b16
kunal-vaishnavi Merge branch 'stack-32528' into stack-32618
96f00c66
kunal-vaishnavi Restrict SparseAttentionIndexer to integer masks
9c8ed58e
kunal-vaishnavi Merge branch 'stack-32526' into stack-32528
7d485c89
kunal-vaishnavi Use integer masks in WebGPU sparse indexer
bc922c25
kunal-vaishnavi Merge branch 'stack-32528' into stack-32618
324128f5
kunal-vaishnavi Hoist QSA query preparation out of block scoring
0e649ff9
kunal-vaishnavi Merge branch 'stack-32526' into stack-32528
75d6b8a8
kunal-vaishnavi Merge branch 'stack-32528' into stack-32618
b150814f
kunal-vaishnavi Support packed QK in CUDA sparse attention indexer
1aec0ffd
kunal-vaishnavi Merge branch 'stack-32526' into stack-32528
dbf7d346
kunal-vaishnavi Support packed QK in WebGPU sparse attention indexer
01dc6e16
kunal-vaishnavi Merge branch 'stack-32528' into stack-32618
7f413610
Copilot Address SparseAttentionIndexer review feedback
ab24124e
Copilot Merge remote-tracking branch 'origin/main' into copilot/implement-spa…
6e6f6471
Copilot Fix SparseAttentionIndexer shape inference test
5def544f
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
d173ebe8
Copilot Address SparseAttentionIndexer review feedback
2217d9a7
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
22e0180c
Copilot Fix packed indexer registrations and docs
34013505
Copilot Fix packed indexer state inference
4e580ccb
Copilot Reduce packed CSA select bindings
7ea6017b
Copilot Regenerate operator documentation
d2bc9bc2
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
b96e68fe
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
01a4a13b
Copilot Fix packed CSA WebGPU uniforms and docs
5550a25b
Copilot Address SparseAttentionIndexer review feedback
c52a153e
apsonawane apsonawane requested a review from copilot-pull-request-reviewer copilot-pull-request-reviewer 26 days ago
copilot-pull-request-reviewer
copilot-pull-request-reviewer commented on 2026-10-01
Copilot Address non-finite CSA softmax review feedback
755e2894
apsonawane
apsonawane commented on 2026-10-01
Copilot Cover positive infinity in CSA softmax tests
2ab40d9d
apsonawane
apsonawane commented on 2026-10-01
Copilot Merge remote-tracking branch 'origin/main' into copilot/implement-spa…
f48290ae
Copilot Fix symbolic shape test formatting
a9237ca3
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
40776e7c
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
7cb834da
Copilot Address packed indexer review feedback
fd809b38
Copilot Merge remote-tracking branch 'origin/main' into copilot/implement-spa…
f1880aa9
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
4a433e9c
Copilot Fix packed indexer CUDA test compilation
761e54be
Copilot Merge remote-tracking branch 'origin/main' into copilot/implement-spa…
c697583f
Copilot Address SparseAttentionIndexer review suggestions
46e741b1
Copilot Merge remote-tracking branch 'origin/main' into copilot/implement-spa…
aff541c3
Copilot Merge latest main and implement CSA review fixes
f5f5b710
Copilot Merge remote-tracking branch 'origin/main' into copilot/implement-spa…
c375a200
Copilot Fix masked CSA candidate selection
47695edb
Copilot Merge remote-tracking branch 'origin/main' into copilot/implement-spa…
885be56f
Copilot Reduce WebGPU int64 visibility test workload
df34d437
Copilot Merge remote-tracking branch 'origin/copilot/implement-sparse-attenti…
d678b0eb
Base automatically changed from copilot/implement-sparse-attention-indexer-again to main 24 days ago
Copilot Merge remote-tracking branch 'origin/main' into copilot/copilotimplem…
662b1236
kunal-vaishnavi kunal-vaishnavi requested a review from apsonawane apsonawane 23 days ago
kunal-vaishnavi
kunal-vaishnavi approved these changes on 2026-10-04
apsonawane
Copilot Merge remote-tracking branch 'origin/main' into copilot/copilotimplem…
c65d910a
Copilot Fix packed sparse indexer review issues
8b9acbb1
Copilot Clarify large-ratio compaction coverage
8dc29988
apsonawane
copilot-swe-agent copilot-swe-agent requested a review from kunal-vaishnavi kunal-vaishnavi 22 days ago
Copilot Merge remote-tracking branch 'origin/main' into copilot/copilotimplem…
120a8838
Copilot Reduce packed indexer WebGPU bindings
01820176
copilot-swe-agent
Copilot Fix non-WebGPU test warning
3eb4630f
Copilot Fix optional output presence inference
8f8aac96
sayanshaw24
sayanshaw24 approved these changes on 2026-10-05
kunal-vaishnavi kunal-vaishnavi merged 3188a2ee into main 21 days ago
kunal-vaishnavi kunal-vaishnavi deleted the copilot/copilotimplement-packed-sparse-attention-indexer branch 21 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
Labels
None yet
Milestone
No milestone
Terms Of Use Privacy Policy Refund Policy Impressum