[Frontend] Introduce Renderer for processing chat messages (using ModelConfig) (#30200)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-01-22 12:44:22 +00:00
committed by GitHub
parent 421012b63a
commit d117a4d1a9
48 changed files with 2141 additions and 1585 deletions
+8 -10
View File
@@ -23,9 +23,10 @@ from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalRegistry
from vllm.outputs import PoolingRequestOutput, RequestOutput
from vllm.plugins.io_processors import get_io_processor
from vllm.pooling_params import PoolingParams
from vllm.renderers import RendererLike
from vllm.sampling_params import SamplingParams
from vllm.tasks import SupportedTask
from vllm.tokenizers import TokenizerLike, cached_tokenizer_from_config
from vllm.tokenizers import TokenizerLike
from vllm.tracing import init_tracer
from vllm.transformers_utils.config import maybe_register_config_serialize_by_value
from vllm.usage.usage_lib import UsageContext
@@ -106,9 +107,7 @@ class AsyncLLM(EngineClient):
"enabling logging without default stat loggers."
)
tokenizer = cached_tokenizer_from_config(self.model_config)
self.input_processor = InputProcessor(self.vllm_config, tokenizer)
self.input_processor = InputProcessor(self.vllm_config)
self.io_processor = get_io_processor(
self.vllm_config,
self.model_config.io_processor_plugin,
@@ -709,13 +708,12 @@ class AsyncLLM(EngineClient):
def tokenizer(self) -> TokenizerLike | None:
return self.input_processor.tokenizer
async def get_tokenizer(self) -> TokenizerLike:
if self.tokenizer is None:
raise ValueError(
"Unable to get tokenizer because `skip_tokenizer_init=True`"
)
def get_tokenizer(self) -> TokenizerLike:
return self.input_processor.get_tokenizer()
return self.tokenizer
@property
def renderer(self) -> RendererLike:
return self.input_processor.renderer
async def is_tracing_enabled(self) -> bool:
return self.observability_config.otlp_traces_endpoint is not None # type: ignore