Higher concurrency limit for gunicorn (#249)

yunfeng-scale · web-flow · commit 04d445ed9d88 · 2023-09-01T15:48:45.000-07:00
diff --git a/model-engine/model_engine_server/api/worker.py b/model-engine/model_engine_server/api/worker.py
@@ -1,8 +1,6 @@
 from uvicorn.workers import UvicornWorker
 
-# The target concurrency is around 50, so we set the limit to 32 with 4 workers
-# for a total concurrency of 128 to allow for some headroom.
-CONCURRENCY_LIMIT = 32
+CONCURRENCY_LIMIT = 1000
 
 
 class LaunchWorker(UvicornWorker):
