refactor(serve): rename generation config helper functions

Rename extract_request_sampling_values to extract_request_gen_config
and merge_sampling_params to merge_gen_config for clearer naming.

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: lvhan028

lmdeploy/serve/core/generation_config.py

@@ -47,11 +47,11 @@ def resolve_default_gen_config(
     return config
 
 
-def merge_sampling_params(
+def merge_gen_config(
     request_values: dict[str, Any],
     default_gen_config: dict[str, Any],
 ) -> dict[str, Any]:
-    """Merge sampling params with request > default_gen_config priority."""
+    """Merge generation config with request > default_gen_config priority."""
     merged: dict[str, Any] = {}
     for key in set(default_gen_config) | set(request_values):
         if key in request_values:
@@ -61,7 +61,7 @@ def merge_sampling_params(
     return merged
 
 
-def extract_request_sampling_values(request: Any) -> dict[str, Any]:
+def extract_request_gen_config(request: Any) -> dict[str, Any]:
     """Extract non-None GenerationConfig fields present on the request."""
     values: dict[str, Any] = {}
     for field in dataclasses.fields(GenerationConfig):
@@ -82,8 +82,8 @@ def build_generation_config(
 ) -> GenerationConfig:
     """Build ``GenerationConfig`` from merged sampling defaults and request
     values."""
-    request_values = extract_request_sampling_values(request)
-    merged = merge_sampling_params(request_values, default_gen_config)
+    request_values = extract_request_gen_config(request)
+    merged = merge_gen_config(request_values, default_gen_config)
     merged.pop('max_new_tokens', None)
     merged.pop('do_sample', None)
     return GenerationConfig(

tests/test_lmdeploy/serve/test_generation_config.py

@@ -4,31 +4,31 @@
 from lmdeploy.messages import GenerationConfig
 from lmdeploy.serve.core.generation_config import (
     build_generation_config,
-    extract_request_sampling_values,
-    merge_sampling_params,
+    extract_request_gen_config,
+    merge_gen_config,
     resolve_default_gen_config,
 )
 from lmdeploy.serve.openai.protocol import ChatCompletionRequest, CompletionRequest
 
 _DEFAULTS = GenerationConfig()
 
 
-def test_merge_sampling_params_priority():
-    merged = merge_sampling_params(
+def test_merge_gen_config_priority():
+    merged = merge_gen_config(
         {'temperature': 0.2},
         {'temperature': 0.5, 'top_k': 10},
     )
     assert merged == {'temperature': 0.2, 'top_k': 10}
 
 
-def test_merge_sampling_params_uses_server_defaults():
-    merged = merge_sampling_params({}, {'temperature': 0.5})
+def test_merge_gen_config_uses_server_defaults():
+    merged = merge_gen_config({}, {'temperature': 0.5})
     assert merged == {'temperature': 0.5}
 
 
-def test_extract_request_sampling_values_only_non_null():
+def test_extract_request_gen_config_only_non_null():
     request = ChatCompletionRequest(model='test', messages='hi', temperature=0.3)
-    values = extract_request_sampling_values(request)
+    values = extract_request_gen_config(request)
     assert values == {'temperature': 0.3}