vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-12 01:54:37 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

Merge branch 'master' into concedo_experimental

Browse source 
# Conflicts:
#	.github/workflows/nix-ci.yml
#	CMakeLists.txt
#	Makefile
#	ggml-cuda.cu
#	ggml-opencl.cpp
#	llama.cpp
This commit is contained in:
Concedo 2024-01-20 23:27:42 +08:00
commit 71e9a64171
43 changed files with 6540 additions and 3500 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/benchmark/benchmark-matmult.cpp
View file

@ -195,7 +195,7 @@ int main(int argc, char ** argv) {
// Set up a the benchmark matrices
// printf("Creating new tensor q11 & Running quantize\n");
struct ggml_tensor * q11 = ggml_new_tensor_2d(ctx, qtype, sizex, sizey);
ggml_quantize_chunk(qtype, (const float *) m11->data, q11->data, 0, nelements, hist_cur.data());
ggml_quantize_chunk(qtype, (const float *) m11->data, q11->data, 0, nelements/m11->ne[0], m11->ne[0], hist_cur.data(), nullptr);
// Set up a the compute graph
// printf("Creating new tensor q31\n");
@ -208,7 +208,7 @@ int main(int argc, char ** argv) {
// Set up a second graph computation to make sure we override the CPU cache lines
// printf("Creating new tensor q12 & Running quantize\n");
struct ggml_tensor * q12 = ggml_new_tensor_2d(ctx, qtype, sizex, sizey);
ggml_quantize_chunk(qtype, (const float *) m12->data, q12->data, 0, nelements, hist_cur.data());
ggml_quantize_chunk(qtype, (const float *) m12->data, q12->data, 0, nelements/m12->ne[0], m12->ne[0], hist_cur.data(), nullptr);
// printf("Creating new tensor q32\n");
struct ggml_tensor * q32 = ggml_mul_mat(ctx, q12, m2);