nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 22 Packages Projects Releases Wiki Activity

server: skip unused log lines on router mode (#24463)

Browse Source
This commit is contained in:
Xuan-Son Nguyen
2026-06-11 11:36:35 +02:00
committed by GitHub
parent 1bfbdb134e
commit 18ef86ecec
1 changed files with 14 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/server/server.cpp
+14 -12
View File
@@ -94,20 +94,22 @@ int llama_server(int argc, char ** argv) {
const bool is_router_server = params.model.path.empty(); const bool is_router_server = params.model.path.empty();
common_params_print_info(params, !is_router_server); common_params_print_info(params, !is_router_server);
// validate batch size for embeddings if (!is_router_server) {
// embeddings require all tokens to be processed in a single ubatch // validate batch size for embeddings
// see https://github.com/ggml-org/llama.cpp/issues/12836 // embeddings require all tokens to be processed in a single ubatch
if (params.embedding && params.n_batch > params.n_ubatch) { // see https://github.com/ggml-org/llama.cpp/issues/12836
SRV_WRN("embeddings enabled with n_batch (%d) > n_ubatch (%d)\n", params.n_batch, params.n_ubatch); if (params.embedding && params.n_batch > params.n_ubatch) {
SRV_WRN("setting n_batch = n_ubatch = %d to avoid assertion failure\n", params.n_ubatch); SRV_WRN("embeddings enabled with n_batch (%d) > n_ubatch (%d)\n", params.n_batch, params.n_ubatch);
params.n_batch = params.n_ubatch; SRV_WRN("setting n_batch = n_ubatch = %d to avoid assertion failure\n", params.n_ubatch);
} params.n_batch = params.n_ubatch;
}
if (params.n_parallel < 0) { if (params.n_parallel < 0) {
SRV_INF("%s", "n_parallel is set to auto, using n_parallel = 4 and kv_unified = true\n"); SRV_INF("%s", "n_parallel is set to auto, using n_parallel = 4 and kv_unified = true\n");
params.n_parallel = 4; params.n_parallel = 4;
params.kv_unified = true; params.kv_unified = true;
}
} }
// for consistency between server router mode and single-model mode, we set the same model name as alias // for consistency between server router mode and single-model mode, we set the same model name as alias