Superminaren/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-04 02:37:27 +02:00
Code Issues Packages Projects Releases Wiki Activity

ggml : fix ggml_backend_buft_get_alloc_size() guard (#28038)

Browse Source
This commit is contained in:
Georgi Gerganov
2026-08-30 20:25:15 +03:00
committed by GitHub
parent 62acc89c26
commit 6d1479c148
2 changed files with 2 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-backend.cpp
+1
View File
@@ -70,6 +70,7 @@ size_t ggml_backend_buft_get_alloc_size(ggml_backend_buffer_type_t buft, const s
// if you hit this assert, update ggml_backend_op_alloc_size_may_expand() accordingly
GGML_ASSERT(size <= ggml_nbytes(tensor) ||
ggml_op_is_empty(tensor->op) ||
ggml_is_quantized(tensor->type) || // [TAG_ALLOC_SIZE_EXPAND]
ggml_backend_op_alloc_size_may_expand(tensor->op));
return size;
ggml/src/ggml-cuda/ggml-cuda.cu
+1
View File
@@ -915,6 +915,7 @@ static size_t ggml_backend_cuda_buffer_type_get_alloc_size(ggml_backend_buffer_t
: ggml_nbytes(tensor);
int64_t ne0 = tensor->ne[0];
// [TAG_ALLOC_SIZE_EXPAND]
if (ggml_is_quantized(tensor->type)) {
if (ne0 % MATRIX_ROW_PADDING != 0) {
GGML_ASSERT(tensor->nb[0] == ggml_element_size(tensor));