From d6280e26bd23a0ce029b12a436b645943e4aeb74 Mon Sep 17 00:00:00 2001
From: Yufeng Li <liyufeng1987@gmail.com>
Date: Fri, 3 May 2024 09:43:53 -0700
Subject: [PATCH] check rotary_embedding with seq length (#20547)

### Description
<!-- Describe your changes. -->
with past/present shared same buffer, the present seq length is
different with total sequence length. The size of cos/sin cache should
be checked with sequence length.


### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
---
 .../contrib_ops/cpu/bert/group_query_attention_helper.h   | 8 ++++----
 .../contrib_ops/cuda/bert/group_query_attention_helper.h  | 8 ++++----
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/onnxruntime/contrib_ops/cpu/bert/group_query_attention_helper.h b/onnxruntime/contrib_ops/cpu/bert/group_query_attention_helper.h
index e2a615eba4..860a4355a3 100644
--- a/onnxruntime/contrib_ops/cpu/bert/group_query_attention_helper.h
+++ b/onnxruntime/contrib_ops/cpu/bert/group_query_attention_helper.h
@@ -192,13 +192,13 @@ Status CheckInputs(const Tensor* query,
                              "head_size shall be a multiple of 16. Got head_size % 16 == ",
                              head_size % 16);
     }
-    if (cos_dims[0] < present_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] < present_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,
diff --git a/onnxruntime/contrib_ops/cuda/bert/group_query_attention_helper.h b/onnxruntime/contrib_ops/cuda/bert/group_query_attention_helper.h
index 1a7c3fcea3..8352397f68 100644
--- a/onnxruntime/contrib_ops/cuda/bert/group_query_attention_helper.h
+++ b/onnxruntime/contrib_ops/cuda/bert/group_query_attention_helper.h
@@ -215,13 +215,13 @@ Status CheckInputs(const Tensor* query,
                              "head_size shall be a multiple of 16. Got head_size % 16 == ",
                              head_size % 16);
     }
-    if (cos_dims[0] < present_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] < present_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,