CUDA: fix MMQ for non-contiguous src0, add tests (llama/10021)

* CUDA: fix MMQ for non-contiguous src0, add tests

* revise test code
This commit is contained in:
Johannes Gäßler
2024-10-24 11:09:36 +02:00
committed by Georgi Gerganov
parent 10eb603a3c
commit ab0385f43b
3 changed files with 13 additions and 11 deletions

View File

@ -3464,7 +3464,7 @@ int64_t ggml_nrows(const struct ggml_tensor * tensor) {
size_t ggml_nbytes(const struct ggml_tensor * tensor) {
size_t nbytes;
size_t blck_size = ggml_blck_size(tensor->type);
const size_t blck_size = ggml_blck_size(tensor->type);
if (blck_size == 1) {
nbytes = ggml_type_size(tensor->type);
for (int i = 0; i < GGML_MAX_DIMS; ++i) {