Merge pull request #312 from janhq/hotfix-high-cpu-usage

tikikun · web-flow · commit 3c5c307a19bc · 2024-01-03T17:22:01.000+07:00
fix high cpu usage
diff --git a/controllers/llamaCPP.h b/controllers/llamaCPP.h
@@ -1537,6 +1537,7 @@ struct llama_server_context {
                 "cache\n");
         kv_cache_clear();
       }
+      std::this_thread::sleep_for(std::chrono::milliseconds(5));
       // TODO: Need to implement queueing using CV for better performance
       // std::unique_lock<std::mutex> lock(mutex_tasks);
       // condition_tasks.wait(lock, [&] { return !queue_tasks.empty(); });

Original file line number	Diff line number	Diff line change
`@@ -1537,6 +1537,7 @@ struct llama_server_context {`
`1537`	`1537`	`"cache\n");`
`1538`	`1538`	`kv_cache_clear();`
`1539`	`1539`	`}`
	`1540`	`+ std::this_thread::sleep_for(std::chrono::milliseconds(5));`
`1540`	`1541`	`// TODO: Need to implement queueing using CV for better performance`
`1541`	`1542`	`// std::unique_lock<std::mutex> lock(mutex_tasks);`
`1542`	`1543`	`// condition_tasks.wait(lock, [&] { return !queue_tasks.empty(); });`