LLM-Inference/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-08-12 22:31:11 +04:00
Code Issues Packages Projects Releases Wiki Activity

adapt to upstream changes

Browse Source
This commit is contained in:
Ruben Ortlam
2026-08-11 14:00:58 +02:00
parent eac4c8b76c
commit f370471abf
2 changed files with 5 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/server/server-models.cpp
+4 -10
View File
@@ -935,7 +935,6 @@ int server_models::can_fit(const buft_memory_map & bmm_req) const {
bmm_total[buft] += mem;
}
}
lru_model_name = sched->pick_victim(lk, "");
}
auto get = [](const buft_memory_map & dmm, ggml_backend_buffer_type_t buft) -> size_t {
@@ -1032,18 +1031,15 @@ buft_memory_map server_models::estimate_model_memory(const std::string & name) {
SRV_INF("estimating memory for model name=%s\n", name.c_str());
std::vector<char *> argv = to_char_ptr_array(child_args);
std::vector<char *> envp = to_char_ptr_array(child_env);
subprocess_s proc;
common_subproc proc;
int options = subprocess_option_no_window | subprocess_option_combined_stdout_stderr;
if (subprocess_create_ex(argv.data(), options, envp.data(), &proc) != 0) {
if (!proc.create(child_args, options, child_env)) {
SRV_ERR("failed to spawn measure process for model name=%s\n", name.c_str());
return {};
}
buft_memory_map result;
FILE * out = subprocess_stdout(&proc);
FILE * out = proc.stdout_file();
if (out) {
char buffer[4096];
while (fgets(buffer, sizeof(buffer), out) != nullptr) {
@@ -1066,9 +1062,7 @@ buft_memory_map server_models::estimate_model_memory(const std::string & name) {
}
}
int exit_code = 0;
subprocess_join(&proc, &exit_code);
subprocess_destroy(&proc);
int exit_code = proc.join();
if (exit_code != 0) {
SRV_ERR("measure process for model name=%s exited with code %d\n", name.c_str(), exit_code);
tools/server/server.cpp
+1 -2
View File
@@ -173,8 +173,7 @@ int llama_server(common_params & params, int argc, char ** argv) {
if (params.measure_only) {
llama_model_params mparams = common_model_params_to_llama(params);
mparams.no_alloc = true;
mparams.use_mmap = false;
mparams.use_mlock = false;
mparams.load_mode = LLAMA_LOAD_MODE_NONE;
llama_model_ptr model{llama_model_load_from_file(params.model.path.c_str(), mparams)};
if (!model) {