happyz
/
gemma.cpp
mirror of https://github.com/google/gemma.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Internal changes 

PiperOrigin-RevId: 836654012
This commit is contained in:
Krzysztof Rymski 2025-11-25 07:05:19 -08:00 committed by Copybara-Service
parent c153d5255b
commit b31e8f98e8
3 changed files with 36 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
gemma/flash_attention.cc
View File

@ -17,7 +17,9 @@
#include <stdint.h>
#include <algorithm>
#include <array>
#include <cmath>
#include <cstdlib>
#include <limits>
#include "compression/types.h" // GEMMA_DISABLED_TARGETS

3
gemma/flash_attention.h
View File

@ -37,7 +37,7 @@ namespace gcpp {
const AttentionActivationsPtrs& activations, ThreadingContext& ctx); \
\
void SingleFlashAttention(size_t start_pos, size_t last_pos, \
const float* HWY_RESTRICT q, \
const BF16* HWY_RESTRICT q, \
const MatPtrT<KV_t>& k, const MatPtrT<KV_t>& v, \
size_t layer_idx, \
const AttentionActivationsPtrs& activations, \
@ -60,6 +60,7 @@ namespace gcpp {
size_t layer_idx, const MatPtr& query_norm_scale, \
AttentionActivationsPtrs& activations, QBatch& qbatch, \
ThreadingContext& ctx); \
\
/* NOLINTNEXTLINE(google-readability-namespace-comments) */ \
} // namespace NAMESPACE

1
ops/ops-inl.h
View File

@ -25,6 +25,7 @@
#include <cstdint>
#include <random>
#include <type_traits> // std::enable_if_t
#include <utility>
#include <vector>
#include "ops/matmul.h"