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

CUDA: enable Gemma FA for HIP/Pascal (#9581)

Browse source
This commit is contained in:
Johannes Gäßler 2024-09-22 09:34:52 +02:00 • committed by GitHub
parent ecd5d6b65b
commit a5b57b08ce
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 12 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

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

@ -3599,7 +3599,7 @@ static bool test_backend(ggml_backend_t backend, test_mode mode, const char * op
if (hs != 128 && logit_softcap != 0.0f) continue;
for (int nh : { 32, }) {
for (int kv : { 512, 1024, }) {
for (int nb : { 1, 2, 4, 8, }) {
for (int nb : { 1, 3, 32, 35, }) {
for (ggml_type type_KV : {GGML_TYPE_F16, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0}) {
test_cases.emplace_back(new test_flash_attn_ext(hs, nh, kv, nb, mask, max_bias, logit_softcap, type_KV));
}