vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

cuda : increase C to 128 for better performance

Browse source
This commit is contained in:
Georgi Gerganov 2024-02-01 16:12:56 +02:00
parent 9a5c2a1681
commit ac26f27028
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
4 changed files with 37 additions and 29 deletions
Download patch file Download diff file Expand all files Collapse all files

2
tests/test-backend-ops.cpp
View file

@ -2210,7 +2210,7 @@ static bool test_backend(ggml_backend_t backend, test_mode mode, const char * op
test_cases.emplace_back(new test_leaky_relu());
#if 1
for (int hs : { 64, 80, 128, }) {
for (int hs : { 128, 64, 80, }) {
for (int nh : { 32, }) {
for (int kv : { 512, 1024, 2048, 4096, }) {
for (int nb : { 1, 2, 4, 8, 512, 1024, 2048, }) {