mirror of
https://github.com/ggerganov/whisper.cpp.git
synced 2025-02-23 10:00:19 +00:00
Fix q_xxs using mul_mat_q (llama/7459)
This commit is contained in:
parent
02d481595b
commit
e98f9ac554
@ -15263,6 +15263,7 @@ static void ggml_sycl_mul_mat(const ggml_tensor * src0, const ggml_tensor * src1
|
|||||||
}
|
}
|
||||||
} else {
|
} else {
|
||||||
bool use_mul_mat_q = min_compute_capability >= VER_4VEC && ggml_is_quantized(src0->type);
|
bool use_mul_mat_q = min_compute_capability >= VER_4VEC && ggml_is_quantized(src0->type);
|
||||||
|
use_mul_mat_q = use_mul_mat_q && (src0->type != GGML_TYPE_IQ2_XXS);
|
||||||
|
|
||||||
if (use_xmx && min_compute_capability >= VER_GEN9 && src1->ne[1] > XMX_MAX_BATCH_SIZE) {
|
if (use_xmx && min_compute_capability >= VER_GEN9 && src1->ne[1] > XMX_MAX_BATCH_SIZE) {
|
||||||
use_mul_mat_q = false;
|
use_mul_mat_q = false;
|
||||||
|
Loading…
x
Reference in New Issue
Block a user