saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-29 20:14:01 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Fix SparseAttention cos/sin cache dimension checks (#20609)

Browse source 
### Description
This PR fixes the dimension checks for the cos/sin caches used in the
rotary embeddings in the `SparseAttention` operator.

### Motivation and Context
This PR ports over the same changes from [this
PR](https://github.com/microsoft/onnxruntime/pull/20547) for
`GroupQueryAttention`.
This commit is contained in:
kunal-vaishnavi 2024-05-08 16:07:02 -07:00 committed by GitHub
parent 58d7b12205
commit 274d162d93
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 4 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
onnxruntime/contrib_ops/cuda/sparse/sparse_attention_helper.h
View file

@ -202,13 +202,13 @@ Status CheckInputs(void* params,
"head_size shall be a multiple of 16. Got head_size = ",
head_size);
}
if (cos_dims[0] < max_sequence_length) {
if (cos_dims[0] < total_sequence_length) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT,
"cos_cache dimension 0 should be of max_sequence_length.");
"cos_cache dimension 0 should be not be less than total_sequence_length.");
}
if (sin_dims[0] < max_sequence_length) {
if (sin_dims[0] < total_sequence_length) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT,
"sin_cache dimension 0 should be of max_sequence_length.");
"sin_cache dimension 0 should be not be less than total_sequence_length.");
}
if (cos_dims[1] > (head_size / 16) * 8 || cos_dims[1] % 8 != 0) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT,