IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[Coverity Scan Issue] fix cs issues in scaled_attn.cpp (#21270)

Browse Source 
* [Coverity Scan Issue] fix cs issues in scaled_attn.cpp

Signed-off-by: Zhai, Xuejun <xuejun.zhai@intel.com>

* Fix review comments

Signed-off-by: Zhai, Xuejun <xuejun.zhai@intel.com>

---------

Signed-off-by: Zhai, Xuejun <xuejun.zhai@intel.com>
This commit is contained in:
Xuejun Zhai 2023-11-27 14:25:02 +08:00 committed by GitHub
parent 72cb4e4820
commit 058001eb84
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 6 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
src/plugins/intel_cpu/src/nodes/scaled_attn.cpp
View File

@ -232,7 +232,7 @@ struct MHAKernel<ScaledDotProductAttention::KT_ONEDNN, T> {
}
PlainTensor<uint8_t> causal_mask;
bool select_nfltmax_at_0; // set attn_score to -FLT_MAX when causal_mask[...] equal to this
bool select_nfltmax_at_0 = false; // set attn_score to -FLT_MAX when causal_mask[...] equal to this
void set_causal_mask(PlainTensor<uint8_t> mask, bool _select_nfltmax_at_0) {
causal_mask = mask;
select_nfltmax_at_0 = _select_nfltmax_at_0;
@ -300,6 +300,7 @@ struct MHAKernel<ScaledDotProductAttention::KT_MLAS, float> {
MHAKernel() {
m_block_size = 4;
select_nfltmax_at_0 = false;
qk_buffers.resize(parallel_get_max_threads(), PlainTensor<float>(true));
}
@ -342,7 +343,9 @@ struct MHAKernel<ScaledDotProductAttention::KT_MLAS, float> {
auto m_blocks = (q_len + m_block_size - 1) / m_block_size;
parallel_for3d(B, H, m_blocks, [&](size_t b, size_t h, size_t m_blk) {
size_t thread_id = static_cast<size_t>(parallel_get_thread_num());
auto thread_id = parallel_get_thread_num();
if (thread_id < 0)
OPENVINO_THROW("The calling thread isn't initialized!");
auto& qk_buf = qk_buffers[thread_id];
auto m_start = m_blk * m_block_size;
@ -450,7 +453,7 @@ struct MHASingleToken {
PlainTensor<float> m_attn_w;
PlainTensor<float> m_temp;
MHASingleToken() : m_attn_w(true), m_temp(true) {}
MHASingleToken() : m_attn_w(true), m_temp(true), select_nfltmax_at_0(false) {}
PlainTensor<uint8_t> causal_mask;
bool select_nfltmax_at_0; // set attn_score to -FLT_MAX when causal_mask[...] equal to this