root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-26 03:14:35 +00:00
Code Issues Actions 7 Packages Projects Releases Wiki Activity

ggml ; remove unused ggml_mul special case

Browse Source 
It would otherwise conflict with the more general
optimization coming with Mamba-2.

* ggml : handle TQ1_0 and TQ2_0 in dequantization-based operators
This commit is contained in:
Francis Couture-Harpin 2024-09-04 13:50:08 -04:00
parent 7f3a619c98
commit 8d61607656
1 changed files with 11 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
ggml/src/ggml.c
View File

@ -9921,6 +9921,8 @@ static void ggml_compute_forward_add(
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
case GGML_TYPE_TQ1_0:
case GGML_TYPE_TQ2_0:
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ3_XXS:
@ -10299,6 +10301,8 @@ static void ggml_compute_forward_add1(
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
case GGML_TYPE_TQ1_0:
case GGML_TYPE_TQ2_0:
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ3_XXS:
@ -10427,6 +10431,8 @@ static void ggml_compute_forward_acc(
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
case GGML_TYPE_TQ1_0:
case GGML_TYPE_TQ2_0:
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ3_XXS:
@ -10562,16 +10568,7 @@ static void ggml_compute_forward_mul_f32(
GGML_ASSERT( nb0 == sizeof(float));
GGML_ASSERT(nb00 == sizeof(float));
if (ggml_nelements(src1) == 1) {
float scale = ((float *) src1->data)[0];
for (int64_t ir = ith; ir < nr; ir += nth) {
if (dst->data != src0->data) {
// src0 is same shape as dst => same indices
memcpy((char *)dst->data + ir*nb1, (char *)src0->data + ir*nb01, ne0 * sizeof(float));
}
ggml_vec_scale_f32(ne0, (float *) ((char *) dst->data + ir*nb1), scale);
}
} else if (nb10 == sizeof(float)) {
if (nb10 == sizeof(float)) {
for (int64_t ir = ith; ir < nr; ir += nth) {
// src0 and dst are same shape => same indices
const int64_t i03 = ir/(ne02*ne01);
@ -13419,6 +13416,8 @@ static void ggml_compute_forward_out_prod(
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
case GGML_TYPE_TQ1_0:
case GGML_TYPE_TQ2_0:
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ3_XXS:
@ -13607,6 +13606,8 @@ static void ggml_compute_forward_set(
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
case GGML_TYPE_TQ1_0:
case GGML_TYPE_TQ2_0:
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ3_XXS: