]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commitdiff
tests: enable kv_unified to prevent cuda oom error on rtx 2060 (#20645)
authorAaron Teo <redacted>
Wed, 18 Mar 2026 09:40:22 +0000 (17:40 +0800)
committerGitHub <redacted>
Wed, 18 Mar 2026 09:40:22 +0000 (17:40 +0800)
Signed-off-by: Aaron Teo <redacted>
tests/test-backend-sampler.cpp

index d4cd62c71e728494860efac54da21f37e1bd987d..58361ae80aea11118597636eac7275fe943fa507 100644 (file)
@@ -89,6 +89,7 @@ struct test_context {
         cparams.n_batch = 512;
         cparams.samplers = configs.data();
         cparams.n_samplers = configs.size();
+        cparams.kv_unified = true;
 
         // If n_seq_max is not specified, calculate it from configs
         if (n_seq_max < 0) {