[Refactor] Pass full VllmConfig to Renderer (#34485)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-02-12 22:48:38 -08:00
committed by GitHub
parent 1b4e8e53f8
commit 2f308214c0
18 changed files with 137 additions and 86 deletions
+6 -1
View File
@@ -59,11 +59,16 @@ class MockModelConfig:
return self.diff_sampling_param or {}
@dataclass
class MockVllmConfig:
model_config: MockModelConfig
def _build_renderer(model_config: MockModelConfig):
_, tokenizer_name, _, kwargs = tokenizer_args_from_config(model_config)
return HfRenderer(
model_config,
MockVllmConfig(model_config),
tokenizer_kwargs={**kwargs, "tokenizer_name": tokenizer_name},
)
@@ -58,6 +58,11 @@ class MockModelConfig:
return self.diff_sampling_param or {}
@dataclass
class MockVllmConfig:
model_config: MockModelConfig
def _build_serving_completion(engine: AsyncLLM) -> OpenAIServingCompletion:
models = OpenAIServingModels(
engine_client=engine,
@@ -74,7 +79,7 @@ def _build_renderer(model_config: MockModelConfig):
_, tokenizer_name, _, kwargs = tokenizer_args_from_config(model_config)
return HfRenderer(
model_config,
MockVllmConfig(model_config),
tokenizer_kwargs={**kwargs, "tokenizer_name": tokenizer_name},
)
@@ -57,6 +57,11 @@ class MockModelConfig:
return self.diff_sampling_param or {}
@dataclass
class MockVllmConfig:
model_config: MockModelConfig
class MockLoRAResolver(LoRAResolver):
async def resolve_lora(
self, base_model_name: str, lora_name: str
@@ -91,7 +96,7 @@ def _build_renderer(model_config: MockModelConfig):
_, tokenizer_name, _, kwargs = tokenizer_args_from_config(model_config)
return HfRenderer(
model_config,
MockVllmConfig(model_config),
tokenizer_kwargs={**kwargs, "tokenizer_name": tokenizer_name},
)
+14 -3
View File
@@ -534,11 +534,16 @@ class MockModelConfig:
return self.diff_sampling_param or {}
@dataclass
class MockVllmConfig:
model_config: MockModelConfig
def _build_renderer(model_config: MockModelConfig):
_, tokenizer_name, _, kwargs = tokenizer_args_from_config(model_config)
return HfRenderer(
model_config,
MockVllmConfig(model_config),
tokenizer_kwargs={**kwargs, "tokenizer_name": tokenizer_name},
)
@@ -749,7 +754,10 @@ async def test_serving_chat_mistral_token_ids_prompt_is_validated():
mock_engine.io_processor = MagicMock()
mock_tokenizer = MagicMock(spec=MistralTokenizer)
mock_renderer = MistralRenderer(mock_engine.model_config, tokenizer_kwargs={})
mock_renderer = MistralRenderer(
MockVllmConfig(mock_engine.model_config),
tokenizer_kwargs={},
)
mock_renderer._tokenizer = mock_tokenizer
# Force the Mistral chat template renderer to return token IDs.
# Choose a prompt length that is < max_model_len, but large enough that
@@ -788,7 +796,10 @@ async def test_serving_chat_mistral_token_ids_prompt_too_long_is_rejected():
mock_engine.io_processor = MagicMock()
mock_tokenizer = MagicMock(spec=MistralTokenizer)
mock_renderer = MistralRenderer(mock_engine.model_config, tokenizer_kwargs={})
mock_renderer = MistralRenderer(
MockVllmConfig(mock_engine.model_config),
tokenizer_kwargs={},
)
mock_renderer._tokenizer = mock_tokenizer
# prompt_token_ids length == max_model_len should be rejected for
# completion-like requests (ChatCompletionRequest).