Distribute and run LLMs with a single file.
C++ cross-platformggufllama-cpplocal-ailocal-inferencelocal-llmopen-source-aisingle-file-executablespeech-to-text
diff --git a/tools/server/server-models-sched.cpp b/tools/server/server-models-sched.cpp
--- a/llama.cpp/tools/server/server-models-sched.cpp
+++ b/llama.cpp/tools/server/server-models-sched.cpp
@@ -277,7 +277,12 @@ void server_models_sched::make_room(const std::string & name) {
};
if (!measured()) {
SRV_INF("waiting for the memory measurement of model name=%s\n", name.c_str());
- models.cv.wait(lk, measured);
+ // Use wait_for() instead of wait() to work around a Cosmopolitan libc bug
+ // where untimed futex waits on XNU (macOS) expire after ~72 minutes,
+ // causing condition_variable::wait() to throw ETIMEDOUT.
+ while (!measured()) {
+ models.cv.wait_for(lk, std::chrono::seconds(30));
+ }
}
}
@@ -298,9 +303,12 @@ void server_models_sched::make_room(const std::string & name) {
// wait for unload to complete
{
std::unique_lock<std::mutex> lk(models.mutex);
- models.cv.wait(lk, [this, &lru_model_name]() {
- return models.mapping[lru_model_name].meta.status == SERVER_MODEL_STATUS_UNLOADED;
- });
+ // Use wait_for() instead of wait() to work around a Cosmopolitan libc
+ // bug where untimed futex waits on XNU (macOS) expire after ~72 minutes,
+ // causing condition_variable::wait() to throw ETIMEDOUT.
+ while (models.mapping[lru_model_name].meta.status != SERVER_MODEL_STATUS_UNLOADED) {
+ models.cv.wait_for(lk, std::chrono::seconds(30));
+ }
}
}
}