fix

.github/workflows/bench_test.yml

@@ -1,9 +1,6 @@
 name: bench_test
 
 on:
-  push:
-    branches:
-    - main
   pull_request:
     branches:
     - main

.github/workflows/e2e_test.yml

@@ -1,9 +1,6 @@
 name: e2e_test
 
 on:
-  push:
-    branches:
-    - main
   pull_request:
     branches:
     - main

.github/workflows/migration_test.yml

@@ -1,9 +1,6 @@
 name: migration_test
 
 on:
-  push:
-    branches:
-    - main
   pull_request:
     branches:
     - main

.github/workflows/offline_inference.yml

@@ -1,9 +1,6 @@
 name: offline_inference
 
 on:
-  push:
-    branches:
-    - main
   pull_request:
     branches:
     - main

.github/workflows/pylint.yml

@@ -1,9 +1,6 @@
 name: pylint
 
 on:
-  push:
-    branches:
-    - main
   pull_request:
     branches:
     - main

.github/workflows/unit_test.yml

@@ -1,9 +1,6 @@
 name: unit_test
 
 on:
-  push:
-    branches:
-    - main
   pull_request:
     branches:
     - main

.github/workflows/whl.yml → .github/workflows/whl_build.yml

@@ -1,9 +1,6 @@
 name: whl_build
 
 on:
-  push:
-    branches:
-    - main
   pull_request:
     branches:
     - main

configs/base.yml

@@ -1,6 +1,7 @@
 SERVER:
   HOST: '127.0.0.1'
   PORT: 37000
+  QUEUE_TYPE: "rayqueue"
 
 RAY:
   RAY_CLUSTER_PORT: 30037

llumnix/config/default.py

@@ -26,6 +26,8 @@
 _C.SERVER.HOST = "localhost"
 # Port number for the server
 _C.SERVER.PORT = 8000
+# Queue type for request output queue
+_C.SERVER.QUEUE_TYPE = "rayqueue"
 # Port number for the request output queue
 _C.SERVER.REQUEST_OUTPUT_QUEUE_PORT = 1234
 # Path to SSL key file for secure connections

llumnix/entrypoints/vllm/api_server.py

@@ -250,7 +250,8 @@ def add_argument(self, *args, **kwargs):
     parser.add_argument('--disable-log-requests-server', action='store_true', help='disable logging requests in server')
     parser.add_argument("--ray-cluster-port", type=int)
     parser.add_argument('--launch-ray-cluster', action='store_true', help='if launch ray cluster in api server')
-    parser.add_argument("--request-output-queue-port", type=int)
+    parser.add_argument("--queue-type", type=str, choices=['rayqueue', 'zmq'], help='queue type for request output queue')
+    parser.add_argument("--request-output-queue-port", type=int, help='port for zeromq')
     parser.add_argument("--config-file", help="path to config file")
     parser = EngineManagerArgs.add_cli_args(parser)