tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-28 12:25:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

imatrix: fix oob writes if src1 is not contiguous (#13286)

Browse Source
This commit is contained in:
Johannes Gäßler
2025-05-04 00:50:37 +02:00
committed by GitHub
parent 36667c8edc
commit 3e959f0976
1 changed files with 8 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
tools/imatrix/imatrix.cpp
View File

@ -46,7 +46,7 @@ private:
common_params m_params; common_params m_params;
std::mutex m_mutex; std::mutex m_mutex;
int m_last_call = 0; int m_last_call = 0;
std::vector<float> m_src1_data; std::vector<char> m_src1_data;
std::vector<char> m_ids; // the expert ids from ggml_mul_mat_id std::vector<char> m_ids; // the expert ids from ggml_mul_mat_id
}; };
@ -93,11 +93,13 @@ bool IMatrixCollector::collect_imatrix(struct ggml_tensor * t, bool ask, void *
const bool is_host = ggml_backend_buffer_is_host(src1->buffer); const bool is_host = ggml_backend_buffer_is_host(src1->buffer);
if (!is_host) { if (!is_host) {
m_src1_data.resize(ggml_nelements(src1)); const size_t src1_nbytes = ggml_nbytes(src1);
ggml_backend_tensor_get(src1, m_src1_data.data(), 0, ggml_nbytes(src1)); m_src1_data.resize(src1_nbytes);
ggml_backend_tensor_get(src1, m_src1_data.data(), 0, src1_nbytes);
} }
const float * data = is_host ? (const float *) src1->data : m_src1_data.data(); const char * data = is_host ? (const char *) src1->data : m_src1_data.data();
GGML_ASSERT(src1->nb[0] == ggml_element_size(src1));
// this has been adapted to the new format of storing merged experts in a single 3d tensor // this has been adapted to the new format of storing merged experts in a single 3d tensor
// ref: https://github.com/ggml-org/llama.cpp/pull/6387 // ref: https://github.com/ggml-org/llama.cpp/pull/6387
@ -144,7 +146,7 @@ bool IMatrixCollector::collect_imatrix(struct ggml_tensor * t, bool ask, void *
const int64_t i11 = idx % src1->ne[1]; const int64_t i11 = idx % src1->ne[1];
const int64_t i12 = row; const int64_t i12 = row;
const float * x = (const float *)((const char *)data + i11*src1->nb[1] + i12*src1->nb[2]); const float * x = (const float *)(data + i11*src1->nb[1] + i12*src1->nb[2]);
for (int j = 0; j < (int)src1->ne[0]; ++j) { for (int j = 0; j < (int)src1->ne[0]; ++j) {
e.values[e_start + j] += x[j]*x[j]; e.values[e_start + j] += x[j]*x[j];
@ -180,7 +182,7 @@ bool IMatrixCollector::collect_imatrix(struct ggml_tensor * t, bool ask, void *
++e.ncall; ++e.ncall;
LOG_DBGV(2, "%s[%d]: %32s, %s, %5d x %5d, %d\n", __func__, m_last_call, wname.c_str(), ggml_op_name(t->op), (int)src1->ne[0], (int)src1->ne[1], (int)src1->type); LOG_DBGV(2, "%s[%d]: %32s, %s, %5d x %5d, %d\n", __func__, m_last_call, wname.c_str(), ggml_op_name(t->op), (int)src1->ne[0], (int)src1->ne[1], (int)src1->type);
for (int row = 0; row < (int)src1->ne[1]; ++row) { for (int row = 0; row < (int)src1->ne[1]; ++row) {
const float * x = data + row * src1->ne[0]; const float * x = (const float *) (data + row * src1->nb[1]);
for (int j = 0; j < (int)src1->ne[0]; ++j) { for (int j = 0; j < (int)src1->ne[0]; ++j) {
e.values[j] += x[j]*x[j]; e.values[j] += x[j]*x[j];
e.counts[j]++; e.counts[j]++;