From 5df183489537a155bbaad9232f25b8e57694d7b8 Mon Sep 17 00:00:00 2001
From: Andy Dai <76841985+Imss27@users.noreply.github.com>
Date: Sat, 5 Oct 2024 10:35:11 -0700
Subject: [PATCH] [Bugfix] Fix order of arguments matters in config.yaml
 (#8960)

---
 .../serving/openai_compatible_server.md       |  2 +-
 tests/data/test_config.yaml                   |  1 +
 tests/test_utils.py                           | 30 ++++++++++++++-----
 vllm/utils.py                                 | 12 +++++++-
 4 files changed, 36 insertions(+), 9 deletions(-)

diff --git a/docs/source/serving/openai_compatible_server.md b/docs/source/serving/openai_compatible_server.md
index 8bb7067faa97c..9132e12a36ba5 100644
--- a/docs/source/serving/openai_compatible_server.md
+++ b/docs/source/serving/openai_compatible_server.md
@@ -140,7 +140,7 @@ $ vllm serve SOME_MODEL --config config.yaml
 ```
 ---
 **NOTE**  
-In case an argument is supplied using command line and the config file, the value from the commandline will take precedence.
+In case an argument is supplied simultaneously using command line and the config file, the value from the commandline will take precedence.
 The order of priorities is `command line > config file values > defaults`.
 
 ---
diff --git a/tests/data/test_config.yaml b/tests/data/test_config.yaml
index 20d499624de2e..42f4f6f7bb992 100644
--- a/tests/data/test_config.yaml
+++ b/tests/data/test_config.yaml
@@ -1,2 +1,3 @@
 port: 12312
+served_model_name: mymodel
 tensor_parallel_size: 2
diff --git a/tests/test_utils.py b/tests/test_utils.py
index c7cb663068c0f..f3017a8582ea8 100644
--- a/tests/test_utils.py
+++ b/tests/test_utils.py
@@ -136,6 +136,8 @@ def parser():
 def parser_with_config():
     parser = FlexibleArgumentParser()
     parser.add_argument('serve')
+    parser.add_argument('model_tag')
+    parser.add_argument('--served-model-name', type=str)
     parser.add_argument('--config', type=str)
     parser.add_argument('--port', type=int)
     parser.add_argument('--tensor-parallel-size', type=int)
@@ -190,33 +192,47 @@ def test_missing_required_argument(parser):
 
 def test_cli_override_to_config(parser_with_config):
     args = parser_with_config.parse_args([
-        'serve', '--config', './data/test_config.yaml',
+        'serve', 'mymodel', '--config', './data/test_config.yaml',
         '--tensor-parallel-size', '3'
     ])
     assert args.tensor_parallel_size == 3
     args = parser_with_config.parse_args([
-        'serve', '--tensor-parallel-size', '3', '--config',
+        'serve', 'mymodel', '--tensor-parallel-size', '3', '--config',
         './data/test_config.yaml'
     ])
     assert args.tensor_parallel_size == 3
+    assert args.port == 12312
+    args = parser_with_config.parse_args([
+        'serve', 'mymodel', '--tensor-parallel-size', '3', '--config',
+        './data/test_config.yaml', '--port', '666'
+    ])
+    assert args.tensor_parallel_size == 3
+    assert args.port == 666
 
 
 def test_config_args(parser_with_config):
     args = parser_with_config.parse_args(
-        ['serve', '--config', './data/test_config.yaml'])
+        ['serve', 'mymodel', '--config', './data/test_config.yaml'])
     assert args.tensor_parallel_size == 2
 
 
 def test_config_file(parser_with_config):
     with pytest.raises(FileNotFoundError):
-        parser_with_config.parse_args(['serve', '--config', 'test_config.yml'])
+        parser_with_config.parse_args(
+            ['serve', 'mymodel', '--config', 'test_config.yml'])
 
     with pytest.raises(ValueError):
         parser_with_config.parse_args(
-            ['serve', '--config', './data/test_config.json'])
+            ['serve', 'mymodel', '--config', './data/test_config.json'])
 
     with pytest.raises(ValueError):
         parser_with_config.parse_args([
-            'serve', '--tensor-parallel-size', '3', '--config', '--batch-size',
-            '32'
+            'serve', 'mymodel', '--tensor-parallel-size', '3', '--config',
+            '--batch-size', '32'
         ])
+
+
+def test_no_model_tag(parser_with_config):
+    with pytest.raises(ValueError):
+        parser_with_config.parse_args(
+            ['serve', '--config', './data/test_config.yaml'])
diff --git a/vllm/utils.py b/vllm/utils.py
index a025c3c40a434..197584867d8b0 100644
--- a/vllm/utils.py
+++ b/vllm/utils.py
@@ -1201,11 +1201,21 @@ def _pull_args_from_config(args: List[str]) -> List[str]:
         config_args = FlexibleArgumentParser._load_config_file(file_path)
 
         # 0th index is for {serve,chat,complete}
+        # followed by model_tag (only for serve)
         # followed by config args
         # followed by rest of cli args.
         # maintaining this order will enforce the precedence
         # of cli > config > defaults
-        args = [args[0]] + config_args + args[1:index] + args[index + 2:]
+        if args[0] == "serve":
+            if index == 1:
+                raise ValueError(
+                    "No model_tag specified! Please check your command-line"
+                    " arguments.")
+            args = [args[0]] + [
+                args[1]
+            ] + config_args + args[2:index] + args[index + 2:]
+        else:
+            args = [args[0]] + config_args + args[1:index] + args[index + 2:]
 
         return args